HK40047016B - Limit of detection based quality control metric - Google Patents

Description

Quality control metrics based on detection limit

Citation Notes

The PCT application form is filed concurrently with this specification and is an integral part of this application. Each claim for a benefit or priority identified in the concurrently filed PCT application form is incorporated herein by reference in its entirety for all purposes.

Background Technology

One of the important tasks in human medical research is the discovery of genetic abnormalities that lead to adverse health consequences. In many cases, specific genes and/or key diagnostic markers have been identified in portions of the genome with abnormal copy numbers. For example, in prenatal diagnosis, extra or missing copies of entire chromosomes are frequently found in genetic disorders. In cancer, copy deletions or duplications of entire chromosomes or chromosomal segments, as well as high-level amplification of specific regions of the genome, often occur.

Most information about copy number variations (CNVs) has been provided at cytogenetic resolution, allowing for the identification of structural abnormalities. Routine procedures for gene screening and biodosimetry have utilized invasive procedures such as amniocentesis, cordocentesis, or chorionic villus sampling (CVS) to obtain cells for chromosome karyotype analysis. Recognizing the need for faster assays that do not require cell culture, fluorescence in situ hybridization (FISH), quantitative real-time PCR (QF-PCR), and array-comparative genomic hybridization (array-CGH) have been developed as molecular cytogenetic methods for analyzing copy number variations.

The advent of technologies that can sequence the entire genome in a relatively short time and the discovery of circulating cell-free DNA (cfDNA) have provided the opportunity to compare genetic material from one chromosome to another without considering the risks associated with invasive sampling methods, and have provided tools for diagnosing multiple copy number variations in the gene sequences of interest.

Due to limitations in existing non-invasive prenatal diagnostic methods, including insufficient sensitivity due to limited levels of cfDNA and technical sequencing biases due to the inherent nature of genomic information, there is a persistent need for non-invasive methods that provide any or all of the specificity, sensitivity, and applicability required to reliably diagnose copy number variations in a variety of clinical settings. Studies have shown that the average length of fetal cfDNA fragments in maternal plasma is shorter than that of maternal cfDNA fragments. This difference between maternal and fetal cfDNA is utilized in the specific implementation described herein to determine CNVs and/or fetal fractions. The implementations disclosed herein address some of these needs. Some implementations provide methods and systems for controlling sample quality in CNV detection, thereby identifying samples with excessively low fetal fractions or read coverage for further processing, such as resequencing. Some implementations provide high analytical sensitivity and specificity for non-invasive prenatal diagnosis and the diagnosis of a variety of diseases.

Summary of the Invention

While the examples herein relate to humans and the language is primarily human-oriented, the concepts described herein apply to the genomes of any plant or animal. These and other objects and features of this disclosure will become more apparent from the following description and the appended claims, or may be learned through practice of this disclosure as described below.

One aspect of this disclosure relates to a method for processing test samples, each test sample containing cell-free nucleic acid fragments derived from the mother and fetus. The method is implemented using a computer system including a memory and one or more processors. In some specific embodiments, the method includes: (a) determining a fetal score of the test sample, wherein the fetal score of the test sample indicates the relative amount of fetal cell-free nucleic acid fragments in the test sample; (b) receiving, by the computer system, sequence reads obtained by sequencing the cell-free nucleic acid fragments in the test sample; (c) aligning, by the computer system, the sequence reads of the cell-free nucleic acid fragments with a reference genome containing a sequence of interest, thereby providing a sequence tag; (d) determining, by the computer system, the coverage of the sequence tag for at least a portion of the reference genome; (e) based on... The sequence tag coverage determined in (d) and the fetal fraction determined in (a) determine whether the test sample is within the exclusion region, wherein the exclusion region is defined by at least a fetal fraction detection limit (LOD) curve, wherein the fetal fraction LOD curve varies with the coverage value and indicates the minimum fetal fraction required to achieve the detection standard given different coverage; and (f) excluding the test sample for determining the CNV of the sequence of interest, or re-sequencing the test sample to obtain a re-sequencing read for determining the CNV of the sequence of interest.

In some implementations, the method also includes, prior to (f), determining that the test sample is negative for the CNV of the sequence of interest.

In some specific implementations, the method also includes: repeating (a)-(d) using re-sequencing sequence reads; determining that the test sample is outside the exclusion region; and classifying the test sample as either a CNV with the sequence of interest or a CNV without the sequence of interest.

In some implementations, the LOD curve for the fetal score is obtained based on the LOD of the affected training samples influenced by CNV. In some implementations, the affected samples include computer-simulated samples. In some implementations, the affected samples include in vitro samples. In some implementations, the affected samples are obtained by combining samples with two or more fetal scores.

In some implementations, the detection criterion is the expected confidence level that the baseline true fetal score is greater than the specified LOD with respect to the observed fetal score. In some implementations, the detection criterion is an X% confidence level that the baseline true fetal score is greater than LOD Y% with respect to the observed fetal score. In some implementations, X is approximately 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, 99%, or 99.5%. In some implementations, Y is a detection confidence level of approximately 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, or 99%. In some implementations, X is 50% and Y is 95%. In some implementations, the specified LOD is determined as the minimum observed fetal score for which Y% of the affected sample can be detected. In some specific implementations, the distribution of the baseline true fetal score at the observation coverage is used to obtain the detection standard for the observed fetal score at that observation coverage.

In some specific implementations, the exclusion region is located below the fetal fractional level of descent (LOD) curve.

In some specific implementations, the exclusion area is defined by the fetal fractional level of distance (LOD) curve and the coverage threshold.

In some specific implementations, the exclusion region is located below both the fetal fractional level of distance (LOD) curve and the coverage threshold.

In some implementations, determining the sequence tag coverage of a portion of the reference genome includes: (i) dividing the reference genome into multiple groups; (ii) determining the number of sequence tags to be aligned with each group; and (iii) using the number of sequence tags in the groups within the portion of the reference genome to determine the sequence tag coverage. In some implementations, the method further includes, prior to (iii), adjusting the number of sequence tags to be aligned with each group by taking into account inter-group variation due to factors other than copy number variation.

In some implementations, the fetal score of a test sample is determined based on the size of the cell-free nucleic acid fragments. In some implementations, the fetal score is determined by: obtaining a frequency distribution of cell-free nucleic acid fragment sizes; and applying this frequency distribution to a model that correlates the fetal score with the frequency of fragment sizes to obtain the fetal score.

In some implementations, the fetal score of the test sample is determined based on the coverage information of the reference genome's groupings. In some implementations, the fetal score is calculated by applying the coverage values of multiple groups of the reference genome to a model that correlates the fetal score with the grouping coverage. In some implementations, multiple groups of the reference genome have higher scores of fetal cell-free nucleic acid fragments than other groups.

In some specific implementations, the fetal score of the test sample is determined based on the coverage information of sex chromosome grouping.

Another aspect of this disclosure relates to a computer system comprising one or more processors and system memory. The one or more processors are configured to perform any of the methods described above.

Additional aspects of this disclosure relate to a computer program product comprising one or more computer-readable non-transitory storage media having computer-executable instructions stored thereon, which, when executed by one or more processors of a computer system, cause the computer system to implement any of the methods described above.

References merged

All patents, patent applications, and other publications mentioned herein, including all sequences disclosed in these references, are expressly incorporated herein by reference to the extent that each individual publication, patent, or patent application is specifically and individually cited and incorporated herein by reference. The relevant portions of all cited references are incorporated herein by reference in their entirety for the purposes indicated in the context of the citation. However, no citation of any reference should be construed as an endorsement of its status as prior art to this disclosure.

Attached Figure Description

Figure 1 schematically illustrates the minimum fetal score required to ensure detection based on coverage.

Figure 2 shows the two-step coverage threshold used to exclude samples.

Figure 3 shows the fetal fraction distribution of the three groups or their samples.

Figure 4A illustrates the workflow 200 for determining CNVs based on some specific implementations of the Limit of Detection (LOD) QC method.

Figure 4B illustrates another LOD QC process for CNV detection.

Figure 5 illustrates empirical and hypothetical data for LOD based on statistical concepts.

Figure 6 shows the detection probability based on fetal score.

Figure 7 illustrates that the estimated fetal score includes errors that cause the estimated fetal score to deviate from the true fetal score.

Figure 8 shows the error in fetal score due to coverage and due to fetal score values.

Figure 9 simulates the actual fetal scores at eight levels of observed fetal scores.

Figure 10 shows the distribution of the true fetal score relative to the observed fetal score at three different levels of error or coverage.

Figure 11 shows the observed fetal scores at 2% and their simulated real fetal score distributions under different errors or coverage conditions.

Figure 12 shows the true fetal score distribution based on the 20th percentile (left) and 25th percentile (right) of the observed fetal score.

Figure 13 shows the LOD, coverage, and observed fetal score.

Figure 14 includes two fetal fractional LOD curves that can be obtained from data similar to that shown in Figure 13.

Figure 15 illustrates a method for determining the presence of copy number variations according to some implementation schemes.

Figure 16A shows a flowchart of the three-path process for evaluating copy number.

Figure 16B illustrates an exemplary process 800 for determining a fetal score based on coverage information according to some specific implementations of this disclosure.

Figure 16C illustrates a specific implementation of the process for determining fetal fractions based on size distribution information.

Figure 16D illustrates an exemplary process 1000 for determining a fetal fraction based on octamer (8-mer) frequency information according to some specific implementations of this disclosure.

Figure 17 illustrates a specific implementation of a dispersion system for generating judgments or diagnoses from test samples.

Figure 18 shows the options for performing various operations at different locations.

Figure 19 shows the chromosome Y coverage (left) and FF fraction estimates (right) of the synthesized samples based on the dilution fraction.

Figure 20 shows the results of linear fitting of LOD contrast coverage.

Figure 21 shows the results of the overlap between predicted LOD and observed LOD.

Figure 22 shows the probability density function of NES.

Figure 23 shows NES coverage based on observed fetal scores.

Figure 24 illustrates data exclusion using exclusion regions defined by LOD curves and read thresholds.

Figure 25 shows the pass rate and failure rate of the first and second runs of the existing method and the LOD QC method.

Figure 26 shows data that were excluded by the two-step threshold method and salvaged by the fetal fraction LOD curve method.

Figure 27 shows a sample rescued using the LOD QC method.

Figure 28 shows the samples excluded by the LOD QC method.

Figure 29 shows the pass rate and failure rate of the existing method and the LOD QC method in two runs.

Figure 30 shows the sample salvaged by the 75% confidence level LOD curve.

Figure 31 illustrates the sensitivity of detecting T21 samples using some specific implementations of the method.

Detailed Implementation

definition

As used in this article, the term “about” for numerical values refers to ±10%.

The term "composed of" means "including and limited to".

The term "consistently made up of" means that a composition, method, or structure may include additional ingredients, steps, and/or portions, provided that such additional ingredients, steps, and/or portions do not substantially alter the essential and novel characteristics of the claimed composition, method, or structure.

Unless otherwise specified, the practices of the methods and systems disclosed herein include conventional techniques and apparatus commonly used in the fields of molecular biology, microbiology, protein purification, protein engineering, protein and DNA sequencing, and recombinant DNA, which are within the scope of the art. Such techniques and apparatus are known to those skilled in the art and are described in numerous texts and references (see, for example, Sambrook et al., “Molecular Cloning: A Laboratory Manual,” 3rd edition (Cold Spring Harbor), [2001]; and Ausubel et al., “Current Protocols in Molecular Biology” [1987]).

Numerical ranges include the numbers that define the range. Each maximum numerical limit given throughout this specification is intended to include every lower numerical limit, as such lower numerical limits are expressly stated herein. Each minimum numerical limit given throughout this specification will include every higher numerical limit, as such higher numerical limits are expressly stated herein. Each numerical range given throughout this specification will include every narrower numerical range falling within such wider numerical ranges, as such narrower numerical ranges are all expressly stated herein.

Unless otherwise defined herein, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Various scientific dictionaries including those containing the terms included herein are well known and available to those skilled in the art. While any methods and materials similar to or equivalent to those described and used herein may also be used in the practice or testing of the embodiments disclosed herein, only a few methods and materials are described herein.

The terms defined below are described more fully with reference to this specification in their entirety. It should be understood that this disclosure is not limited to the specific methods, procedures, and reagents described, as these methods, procedures, and reagents may vary depending on the context in which they are used by those skilled in the art. As used herein, unless the context clearly indicates otherwise, the singular terms “an,” “a,” and “the” include plural references.

Unless otherwise specified, nucleic acids are written from left to right with the 5' to 3' orientation, and the amino acid sequence is written from left to right with the amino to carboxyl orientation.

The limit of detection (LOD) is the minimum level of signal (e.g., analyte, fetal score, score of indicative condition, etc.) that can be defined with confidence for detection. In this application, LOD is the minimum level of fetal score (or other analyte) required to define the confidence level for detecting aneuploidy/CNV.

As used in this paper, the term "parameter" refers to a physical characteristic whose value or other property affects related conditions (such as copy number variation). In some cases, the term parameter refers to a variable that affects a mathematical relationship or model output, which can be an independent variable (i.e., the model's input) or an intermediate variable based on one or more independent variables. Depending on the scope of the model, the output of one model can become the input of another model, thus becoming a parameter of that other model.

The term "fragment size parameter" refers to a parameter relating to the size or length of a fragment or set of fragments (such as nucleic acid fragments); for example, cfDNA fragments obtained from bodily fluids. As used herein, a parameter is "biased toward fragment size or size range" when: 1) the fragment size or size range is advantageously weighted, for example, when a count associated with a fragment of that size or size range is given a greater weight than counts of other sizes or ranges; or 2) the parameter is obtained from values that are advantageously weighted toward fragment size or size range, for example, a ratio obtained from counts with greater weight when associated with a fragment of that size or size range. Fragment size or size range can be a characteristic of the genome or a portion thereof when the genome produces nucleic acid fragments of a size or size range that are richer or have a higher concentration than nucleic acid fragments from another genome or another part of the same genome.

The term "weighting" refers to modifying a quantity (such as a parameter or variable) using one or more values or functions considered as "weights". In some implementations, the parameter or variable is multiplied by the weights. In other implementations, the parameter or variable is modified exponentially. In some implementations, the function can be linear or nonlinear. Examples of suitable nonlinear functions include, but are not limited to, the Heaviside step function, the boxcar function, the step function, or the sigmoid function. Weighting the original parameter or variable systematically increases or decreases the value of the weighted variable. In various implementations, weighting can produce positive, non-negative, or negative values.

In this document, the term "copy number variation" refers to the copy number variation of a nucleic acid sequence present in a test sample compared to the copy number of the nucleic acid sequence present in a reference sample. In some implementations, the nucleic acid sequence is 1 kb or larger. In some cases, the nucleic acid sequence is the entire chromosome or a significant portion thereof. A "copy number variant" is a nucleic acid sequence whose copy number difference is detected by comparing the level of the nucleic acid sequence of interest in the test sample to the expected level of the nucleic acid sequence of interest. For example, comparing the level of the nucleic acid sequence of interest in the test sample to the level of the nucleic acid sequence present in a qualified sample. Copy number variants/variations include deletions (including microdeletions), insertions (including microinsertions), duplications, doubling, and translocations. CNVs cover chromosomal aneuploidy and partial aneuploidy.

In this article, the term "aneuploidy" refers to an imbalance of genetic material caused by the loss or gain of a whole chromosome or a part of a chromosome.

In this article, the terms "chromosomal aneuploidy" and "complete chromosomal aneuploidy" refer to an imbalance of genetic material caused by the loss or gain of a complete chromosome, and include germline aneuploidy and mosaic aneuploidy.

In this article, the terms “partial aneuploidy” and “partial chromosomal aneuploidy” refer to an imbalance of genetic material caused by the loss or gain of a portion of a chromosome (e.g., partial haploidy and partial triploidy), and encompass imbalances caused by translocations, deletions, and insertions.

The term "multiple" means more than one element. For example, as used herein, it refers to multiple nucleic acid molecules or sequence tags sufficient to identify significant differences in copy number variation between test samples and qualified samples using the methods disclosed herein. In some embodiments, each test sample yields at least about 3 × ^10⁶ sequence tags between about 20 bp and 40 bp. In some exemplary embodiments, each test sample provides data for at least about 5 × ^10⁶ , 8 × ^10⁶ , 10 × ^10⁶ , 15 × ^10⁶ , 20 × ^10⁶ , 30 × ^10⁶ , 40 × ^10⁶ , or 50 × ^10⁶ sequence tags, each with a length between about 20 bp and 40 bp.

The term "paired-end read" refers to a read obtained from each end of a nucleic acid fragment through paired-end sequencing. Paired-end sequencing may involve fragmenting a polynucleotide chain into short sequences called insert sequences. Fragmentation is optional or unnecessary for relatively short polynucleotides, such as cell-free DNA molecules.

The terms "polynucleotide," "nucleic acid," and "nucleic acid molecule" are used interchangeably and refer to a sequence of covalently linked nucleotides (i.e., ribonucleotides of RNA and deoxyribonucleotides of DNA), in which the 3' position of the pentose sugar of one nucleotide is linked to the 5' position of the pentose sugar of the next nucleotide via a phosphodiester group. Nucleotides include sequences of any form of nucleic acid, including but not limited to RNA and DNA molecules (such as cfDNA molecules). The term "polynucleotide" includes, but is not limited to, single-stranded and double-stranded polynucleotides.

In this document, the term "test sample" refers to a sample typically derived from a biological fluid, cell, tissue, organ, or organism and containing nucleic acids or mixtures thereof, which contain at least one nucleic acid sequence for which a copy number variation is to be screened. In some embodiments, the sample contains at least one nucleic acid sequence for which a copy number variation is suspected to have occurred. Such samples include, but are not limited to, sputum/oral fluid, amniotic fluid, blood, blood fractions or fine-needle biopsy samples (e.g., surgical biopsy, fine-needle biopsy, etc.), urine, peritoneal fluid, pleural fluid, etc. Although samples are typically taken from human subjects (e.g., patients), the determination can be used for copy number variation (CNV) in any mammal, including but not limited to dogs, cats, horses, goats, sheep, cattle, pigs, etc. Samples may be used directly as is from the biological source or after pretreatment to alter the properties of the sample. For example, such pretreatment may include preparing plasma from blood, diluting viscous fluids, etc. Pretreatment methods also include, but are not limited to, filtration, precipitation, dilution, distillation, mixing, centrifugation, freezing, lyophilization, concentration, amplification, nucleic acid fragmentation, inactivation of interfering components, addition of reagents, lysis, etc. If such pretreatment methods are used on samples, these methods typically result in the retention of the nucleic acid of interest in the test sample, sometimes at a concentration proportional to that in the untreated test sample (e.g., a sample not subjected to any such pretreatment methods). For the purposes of the methods described herein, such “treated” or “processed” samples are still considered biological “test” samples.

In this document, the terms "qualified sample" or "unaffected sample" refer to a sample containing a mixture of nucleic acids present at a known copy number (compared to the nucleic acids in the test sample), and a qualified or unaffected sample is a sample that is normal with respect to the nucleic acid sequence of interest, i.e., not aneuploid. In some embodiments, qualified samples are used as unaffected training samples in a training set to derive sequence masking or sequence maps. In some embodiments, qualified samples are used to identify one or more normalized chromosomes or regions of the chromosome under consideration. For example, qualified samples can be used to identify the normalized chromosome of chromosome 21. In this case, the qualified sample is a sample that is not a 21-triploid sample. Another example includes qualified samples using only female individuals as the X chromosome. Qualified samples can also be used for other purposes, such as determining thresholds for identifying affected samples, identifying thresholds for defining masked regions on a reference sequence, determining expected coverage measures for different regions of the genome, etc.

In this document, the term "training set" refers to a set of training samples that may include affected and/or unaffected samples and are used to develop models for analyzing test samples. In some embodiments, the training set includes unaffected samples. In these embodiments, a training set of samples unaffected by the copy number variation of interest is used to establish a threshold for determining CNVs. Unaffected samples in the training set can be used as qualifying samples to identify normalized sequences (e.g., normalized chromosomes), and the chromosome dose of the unaffected samples is used to set a threshold for each of the sequences of interest (e.g., the chromosome of interest). In some embodiments, the training set includes affected samples. Affected samples in the training set can be used to confirm that affected test samples are easily distinguishable from unaffected samples.

The training set is also a statistical sample of the population of interest, which will not be confused with biological samples. A statistical sample typically includes multiple individuals whose data are used to determine one or more quantitative values of interest applicable to the population. A statistical sample is a subset of individuals in the population of interest. These individuals can be people, animals, tissues, cells, other biological samples (i.e., a statistical sample may include multiple biological samples), and other individual entities that provide data points for statistical analysis.

Typically, training and validation sets are used together. The term "validation set" refers to a set of individuals in a statistical sample whose data are used to validate or evaluate the quantitative values of interest determined using the training set. In some implementations, for example, the training set provides data for computing a mask for a reference sequence, while the validation set provides data for evaluating the effectiveness or impact of the mask.

As used herein, "copy number assessment" refers to a statistical assessment of the gene sequence status in relation to the copy number of the sequence. For example, in some embodiments, the assessment includes determining the presence of a gene sequence. In some embodiments, the assessment includes identifying partial or complete aneuploidy of a gene sequence. In other embodiments, the assessment includes distinguishing two or more samples based on the copy number of the gene sequence. In some embodiments, the assessment includes performing statistical analyses, such as normalization and comparisons, based on the copy number of the gene sequence.

The term "qualified nucleic acid" is used interchangeably with "qualified sequence," which is the sequence of interest or nucleic acid against which the quantity is compared. A qualified sequence is a sequence that is preferably present in a biological sample in a known manner (i.e., the quantity of the qualified sequence is known). Generally, a qualified sequence is a sequence present in a "qualified sample." A "qualified sequence of interest" is a qualified sequence whose quantity is known in a qualified sample and is associated with differences in the sequence of interest between control subjects and individuals with a medical condition.

In this document, the terms "sequence of interest" or "nucleic acid sequence of interest" refer to nucleic acid sequences associated with differences in sequence representation between healthy and diseased individuals. A sequence of interest can be a sequence on a chromosome that is misrepresented (i.e., overrepresented or underrepresented) in a disease or genetic condition. A sequence of interest can be a portion of a chromosome (i.e., a chromosomal segment) or the entire chromosome. For example, a sequence of interest can be a chromosome overrepresented in aneuploidy or a gene encoding a tumor suppressor factor underrepresented in cancer. Sequences of interest include overrepresented or underrepresented sequences in the total or subpopulation of cells in a subject. A "qualified sequence of interest" is a sequence of interest in a qualified sample. A "test sequence of interest" is a sequence of interest in a test sample.

The term "normalized sequence" as used herein refers to a sequence used to normalize the number of sequence tags mapped to the sequence of interest associated with the normalized sequence. In some embodiments, the normalized sequence comprises a robust chromosome. A "robust chromosome" is a chromosome that is unlikely to be aneuploid. In some cases involving human chromosomes, a robust chromosome is any chromosome other than the X chromosome, Y chromosome, chromosome 13, chromosome 18, and chromosome 21. In some embodiments, the normalized sequence shows the variability in the number of sequence tags mapped to it between samples and sequencing runs, which approximates the variability of the sequence of interest used as a normalization parameter. The normalized sequence can distinguish an affected sample from one or more unaffected samples. In some specific embodiments, the normalized sequence optimally or effectively distinguishes an affected sample from one or more unaffected samples when compared with other potential normalized sequences, such as other chromosomes. In some embodiments, the variability of the normalized sequence is calculated based on the chromosomal dose variability of the sequence of interest between samples and sequencing runs. In some embodiments, the normalized sequence is identified in a set of unaffected samples.

"Normalized chromosome," "normalized denominator chromosome," or "normalized chromosome sequence" are examples of a "normalized sequence." A "normalized chromosome sequence" may consist of a single chromosome or a group of chromosomes. In some embodiments, a normalized sequence contains two or more robust chromosomes. In some embodiments, robust chromosomes are all chromosomes other than the X chromosome, Y chromosome, chromosome 13, chromosome 18, and chromosome 21. A "normalized segment" is another example of a "normalized sequence." A "normalized segment sequence" may consist of a single segment of a chromosome or may consist of two or more segments of the same or different chromosomes. In some embodiments, a normalized sequence is intended to normalize variability such as process-related variability, inter-chromosomal (intra-run) variability, and inter-sequencing (inter-run) variability.

In this paper, the term "discriminability" refers to the property of a normalized chromosome that enables one or more unaffected (i.e., normal) samples to be distinguished from one or more affected (i.e., aneuploid) samples. The normalized chromosome exhibiting the greatest "discriminability" is a chromosome or set of chromosomes that provides the largest statistical difference between the chromosome dose of the chromosome of interest in a qualified sample set and the distribution of the chromosome dose of the same chromosome of interest in the corresponding chromosomes of one or more affected samples.

In this paper, the term "variability" refers to the characteristics of a normalized chromosome that enable the differentiation of one or more unaffected (i.e., normal) samples from one or more affected (i.e., aneuploid) samples. The variability of a normalized chromosome measured in a qualified sample set refers to the variability of the number of sequence tags mapped to it, which approximates the variability of the number of sequence tags mapped to the chromosome of interest used as a normalization parameter.

In this article, the term "sequence tag density" refers to the number of sequence reads mapped to a reference genome sequence. For example, the sequence tag density of chromosome 21 is the number of sequence reads of chromosome 21 mapped to the reference genome generated by sequencing methods.

In this article, the term "sequence tag density ratio" refers to the ratio of the number of sequence tags mapped to a chromosome (e.g., chromosome 21) of the reference genome to the length of the reference genome chromosome.

In this document, the term "sequence dose" refers to a parameter that correlates the number of sequence tags or another parameter of the identified sequence of interest with the number of sequence tags or another parameter of the identified normalized sequence. In some cases, sequence dose is the ratio of the sequence tag coverage or other parameter of the sequence of interest to the sequence tag coverage or other parameter of the normalized sequence. In some cases, sequence dose refers to a parameter that correlates the sequence tag density of the sequence of interest with the sequence tag density of the normalized sequence. "Test sequence dose" is a parameter that correlates the sequence tag density or other parameter of the sequence of interest (e.g., chromosome 21) identified in a test sample with the sequence tag density or other parameter of the normalized sequence (e.g., chromosome 9). Similarly, "qualified sequence dose" is a parameter that correlates the sequence tag density or other parameter of the sequence of interest identified in a qualified sample with the sequence tag density or other parameter of the normalized sequence.

The term "coverage" refers to the abundance of sequence tags mapped to a defined sequence. Coverage can be quantitatively expressed by sequence tag density (or sequence tag count), sequence tag density ratio, normalized coverage metric, adjusted coverage value, etc.

The term "coverage metric" refers to a modification of the original coverage and is typically expressed as the relative amount (sometimes called count) of sequence tags in a genomic region, such as in a group. Coverage metrics can be obtained by normalizing, adjusting, and/or correcting the original coverage or count of a genomic region. For example, a normalized coverage metric for a region can be obtained by dividing the count of sequence tags mapped to the region by the total number of sequence tags mapped to the entire genome. Normalized coverage metrics allow for comparison of group coverage across different samples, which may have different sequencing depths. Normalized coverage metrics differ from sequence doses in that the latter are typically obtained by dividing by the count of tags mapped to a subset of the entire genome. This subset is one or more normalized segments or chromosomes. Regardless of normalization, coverage metrics can be corrected for global map variations between regions on the genome, G-C score variations, robust chromosomal outliers, etc.

In this article, the term "next-generation sequencing (NGS)" refers to sequencing methods that allow for massively parallel sequencing of cloned amplified molecules and individual nucleic acid molecules. Non-limiting examples of NGS include sequencing-by-ligation and sequencing-by-synthesis using reversible dye terminators.

In this paper, the term "parameter" refers to a numerical value that characterizes a system property. Typically, a parameter numerically represents the numerical relationship between quantitative datasets and/or between quantitative datasets. For example, the ratio (or a function of the ratio) between the number of sequence tags mapped to a chromosome and the length of the chromosome to which the tags are mapped is a parameter.

In this document, the terms "threshold" and "qualification threshold" refer to any number used as a cutoff value to characterize a sample (such as a test sample containing nucleic acids from an organism suspected of having a medical condition). A threshold can be compared to a parameter value to determine whether a sample producing that parameter value indicates a medical condition in the organism. In some embodiments, a qualified dataset is used to calculate the qualification threshold, and this qualification threshold is used as a diagnostic limit for copy number variations (e.g., aneuploidy) in the organism. If the result obtained from the methods disclosed herein exceeds the threshold, the subject can be diagnosed with a copy number variation, such as triploidy 21. An appropriate threshold for the methods described herein can be determined by analyzing normalized values (e.g., chromosomal dose, NCV, or NSV) of the calculated sample training set. A threshold can be determined using qualified (i.e., unaffected) samples from a training set that includes both qualified (i.e., unaffected) samples and affected samples. Samples in the training set known to have chromosomal aneuploidy (i.e., affected samples) can be used to confirm that the selected threshold can be used to distinguish affected samples from unaffected samples in the test set (see embodiments herein). The choice of threshold depends on the level of confidence the user wishes to be able to classify. In some implementations, the training set used to determine the appropriate threshold includes at least 10, at least 20, at least 30, at least 40, at least 50, at least 60, at least 70, at least 80, at least 90, at least 100, at least 200, at least 300, at least 400, at least 500, at least 600, at least 700, at least 800, at least 900, at least 1000, at least 2000, at least 3000, at least 4000, or more qualified samples. Using a larger set of qualified samples can be advantageous to improve the diagnostic utility of the threshold.

The term "group" refers to a segment of sequence or a segment of the genome. In some implementations, groups are adjacent to each other within the genome or chromosome. Each group may define a nucleotide sequence within a reference genome. Depending on the specific application and the analysis required for sequence tag density, the size of the group can be 1 kb, 100 kb, 1 Mb, etc. In addition to its location within the reference sequence, the group may also have other characteristics, such as sample coverage and sequence structure characteristics (such as G-C score).

As used herein, the term "masking threshold" refers to a quantity compared based on the number of sequence tags in a sequence group, where groups with values exceeding the masking threshold are masked. In some embodiments, the masking threshold may be a percentile, absolute count, mapping quality score, or other suitable value. In some embodiments, the masking threshold may be defined as a percentile of the coefficient of variation among multiple unaffected samples. In other embodiments, the masking threshold may be defined as a mapping quality score (e.g., MapQ score), which is associated with the reliability of aligning a sequence read to a reference genome. It should be noted that the masking threshold differs from the copy number variation (CNV) threshold, which is a cutoff value characterizing samples containing nucleic acids from organisms suspected of having a CNV-related medical condition. In some embodiments, the CNV threshold is defined relative to the normalized chromosome value (NCV) or normalized segment value (NSV) described elsewhere herein.

In this document, the term "normalized value" refers to a numerical value that correlates the number of sequence tags of the identified sequence of interest (e.g., chromosome or chromosome segment) with the number of sequence tags of the identified normalized sequence (e.g., normalized chromosome or normalized chromosome region). For example, a "normalized value" may be a chromosome dose as described elsewhere in this document, or it may be an NCV, or it may be an NSV as described elsewhere in this document.

The term "read" refers to a sequence obtained from a portion of a nucleic acid sample. Typically, though not strictly necessary, a read represents a short sequence of adjacent base pairs in the sample. Reads can be symbolically represented by the base pair sequence (A, T, C, or G) of the sample portion. Reads can be stored in a storage device and processed as needed to determine whether the read matches a reference sequence or meets other criteria. Reads can be obtained directly from a sequencing device or indirectly from stored sequence information about the sample. In some cases, reads are DNA sequences of sufficient length (e.g., at least about 25 bp) that can be used to identify larger sequences or regions, for example, that can be aligned and specifically assigned to chromosomal or genomic regions or genes.

The term "genome read" is used to refer to any segment of an individual's whole genome.

In this document, the term "sequence tag" may be used interchangeably with the term "mapped sequence tag," referring to a sequence read that has been specifically assigned (i.e., mapped) to a larger sequence (e.g., a reference genome) through alignment. Mapped sequence tags are uniquely mapped to the reference genome; that is, they are assigned to a single location within the reference genome. Unless otherwise specified, tags on the same sequence mapped to the reference sequence are counted once. Tags may be provided by data structures or other collections of data. In some embodiments, the tag contains the read sequence and information associated with that read, such as the sequence's location in the genome, e.g., its location on a chromosome. In some embodiments, the location is specific to the positive strand orientation. In alignments with the reference genome, tags may be defined to allow a limited number of mismatches. In some embodiments, tags that can be mapped to more than one location on the reference genome, i.e., tags that are not uniquely mapped, may not be included in the analysis.

The term "non-redundant sequence tag" refers to a sequence tag that does not map to the same locus, which is counted in some implementations to determine the normalized chromosome value (NCV). Sometimes, multiple sequence reads are aligned to the same location on a reference genome, resulting in redundant or repetitive sequence tags. In some implementations, repetitive sequence tags mapped to the same location are omitted or counted as a single "non-redundant sequence tag" to determine the NCV. In some implementations, non-redundant sequence tags aligned to non-excluded loci are counted to produce a "non-excluded locus count" (NES count) for determining the NCV.

The term "site" refers to a unique location on a reference genome (i.e., chromosome ID, chromosome location, and orientation). In some implementations, a site can provide the location of a residue, sequence tag, or segment on the sequence.

"Exclusion sites" are sites located in regions of the reference genome that have been excluded from the counting of sequence tags. In some implementations, exclusion sites are located in chromosomal regions containing repetitive sequences (e.g., centromeres and telomeres) and chromosomal regions shared by more than one chromosome (e.g., regions on the Y chromosome that are also on the X chromosome).

"Non-exclusion sites" (NSEs) are sites in the reference genome that were not excluded for counting sequence tags.

The “Non-Exclusion Site Count” (NES count) is the number of sequence tags mapped to NSEs on the reference genome. In some implementations, the NES count is the number of non-redundant sequence tags mapped to NESs. In some implementations, coverage and related parameters (such as normalized coverage metrics), global map removal coverage metrics, and chromosome dose are based on the NES count. In one example, chromosome dose is calculated as the ratio of the NES count of the chromosome of interest to the normalized chromosome count.

The Normalized Chromosome Value (NCV) correlates the coverage of the test sample with the coverage of the training/qualified sample set. In some implementations, the NCV is based on chromosome dose. In some implementations, the NCV correlates the difference between the chromosome dose of the chromosome of interest in the test sample and the average of the corresponding chromosome doses in the qualified sample set, and can be calculated as follows:

Where and are the estimated mean and standard deviation of the dose of the j-th chromosome in the qualified sample set, respectively, and x _{_ij} is the ratio (dose) of the j-th chromosome in the observed test sample i.

In some implementations, NCV can be calculated "on the fly" as follows: by correlating the chromosome dose of the chromosome of interest in the test sample with the median chromosome dose of the corresponding chromosome in multiplex samples sequenced on the same flow cell.

Where M _{_j} is the estimated median dose of the j-th chromosome in a multiplex sample set sequenced on the same flow cell; is the standard deviation of the dose of the j-th chromosome in one or more multiplex sample sets sequenced on one or more flow cells, and x _{_ij} is the observed dose of the j-th chromosome in test sample i. In this embodiment, test sample i is one of the multiplex samples sequenced on the same flow cell where M_{_j} was determined.

For example, for chromosome 21 of interest in test sample A, which is sequenced in one of 64 multiplex samples in a flow-through pool, the NCV of chromosome 21 in test sample A is calculated as follows: the dose of chromosome 21 in sample A minus the median dose of chromosome 21 determined in the 64 multiplex samples, divided by the standard deviation of the dose of chromosome 21 determined in the 64 multiplex samples in flow-through pool 1 or another flow-through pool.

As used herein, the term "aligned, alignment, or alignment" refers to the process of comparing a read or tag with a reference sequence to determine whether the reference sequence contains the read sequence. If the reference sequence contains the read, the read may be mapped to the reference sequence, or in some embodiments, to a specific location within the reference sequence. In some cases, alignment simply indicates whether a read is a member of a particular reference sequence (i.e., whether the read is present in the reference sequence). For example, aligning a read with a reference sequence of human chromosome 13 will indicate whether the read is present in the reference sequence of chromosome 13. The tool providing this information may be called a setmembership tester. In some cases, alignment additionally indicates the location in the reference sequence to which the read or tag is mapped. For example, if the reference sequence is a human whole genome sequence, alignment may indicate that the read is present on chromosome 13, and may also indicate that the read is present on a specific strand and/or site on chromosome 13.

An aligned read or tag is one or more sequences identified, in terms of their nucleic acid molecular sequence, as matching a known sequence from a reference genome. Alignment can be performed manually, although it is typically implemented using computer algorithms because it is not possible to align reads within a reasonable timeframe for implementing the methods disclosed herein. An example of an algorithm derived from alignment sequences is the High Efficiency Local Alignment of Nucleotide Data (ELAND) computer program, distributed as part of the Illumina Genomics analysis pipeline. Alternatively, Bloom filters or similar set membership testers can be used to align reads to a reference genome. See U.S. Patent Application 61/552,374, filed October 27, 2011, the entire contents of which are incorporated herein by reference. The matching of sequence reads in an alignment can be 100% sequence matching or less than 100% (incomplete matching).

As used in this article, the term "mapping" refers to the specific assignment of sequence reads to larger sequences, such as a reference genome, through alignment.

As used herein, the terms “reference genome” or “reference sequence” refer to any specific known genome sequence of any organism or virus, whether partial or complete, that can be used as a reference for an identification sequence from a subject. For example, reference genomes for human subjects and many other organisms can be found at the National Center for Biotechnology Information (ncbi.nlm.nih.gov). A “genome” refers to the complete genetic information of an organism or virus expressed as a nucleic acid sequence.

In various implementations, the reference sequence is significantly larger than the read it is aligned with. For example, the reference sequence may be at least about 100 times larger than the aligned read, or at least about 1,000 times larger, or at least about 10,000 times larger, or at least about ^10⁵ times larger, or at least about ^10⁶ times larger, or at least about ^10⁷ times larger.

In one example, the reference sequence is the sequence of the full-length human genome. Such a sequence may be called a genomic reference sequence. In another example, the reference sequence is limited to a specific human chromosome, such as chromosome 13. In some implementations, the reference chromosome is a Y chromosome sequence from human genome version hg19. Such a sequence may be called a chromosomal reference sequence. Other examples of reference sequences include genomes of other species, as well as chromosomes, subchromosomal regions (such as chains), etc., of any species.

In various implementations, the reference sequence is a shared sequence or other combination derived from multiple individuals. However, in some applications, the reference sequence may be taken from a specific individual.

In this article, the term "clinically relevant sequence" refers to a nucleic acid sequence that is known or suspected to be associated with or related to a genetic or disease condition. Determining the presence of a clinically relevant sequence can be used to confirm or validate a medical diagnosis, or to provide prognostic information for disease progression.

When used in the context of nucleic acids or mixtures of nucleic acids, the term "derived" herein refers to the manner in which the nucleic acids are obtained from their source. For example, in one embodiment, a mixture of nucleic acids derived from two different genomes means that the nucleic acids (e.g., cfDNA) are naturally released by the cells through a naturally occurring process, such as necrosis or apoptosis. In another embodiment, a mixture of nucleic acids derived from two different genomes means that the nucleic acids were extracted from two different types of cells from a subject.

When used in the context of obtaining a specific quantitative value, the term "based on" in this paper means using another quantity as input to calculate a specific quantitative value as output.

The term "patient sample" as used herein refers to a biological sample obtained from a patient, i.e., a recipient of medical attention, care, or treatment. A patient sample can be any sample described herein. In some embodiments, the patient sample is obtained non-invasively, such as through peripheral blood or fecal samples. The methods described herein are not necessarily limited to humans. Therefore, various veterinary applications are considered, in which case the patient sample can be a sample from a non-human mammal (e.g., a feline, pig, horse, cattle, etc.).

In this article, the term "mixed sample" refers to a sample containing a mixture of nucleic acids derived from different genomes.

In this article, the term "maternal sample" refers to a biological sample obtained from a pregnant subject (e.g., a woman).

In this document, the term "biofluid" refers to a liquid derived from a biological source and includes, for example, blood, serum, plasma, sputum, lavage fluid, cerebrospinal fluid, urine, semen, sweat, tears, saliva, etc. As used herein, the terms "blood," "plasma," and "serum" explicitly encompass their fraction or processing portion. Similarly, in cases where the sample is derived from a biopsy, swab, smear, etc., "sample" explicitly encompasses the processing fraction or portion derived from the biopsy, swab, smear, etc.

In this article, the terms "maternal nucleic acid" and "fetal nucleic acid" refer to the nucleic acid of the pregnant female subject and the nucleic acid of the fetus she is carrying, respectively.

As used in this article, the term “corresponding to” sometimes refers to the fact that nucleic acid sequences (e.g., genes or chromosomes) present in the genomes of different subjects do not necessarily have the same sequence in all genomes, but are used to provide the identity of the sequence of interest (e.g., genes or chromosomes) rather than genetic information.

As used herein, the term "fetal fraction" refers to the fraction of fetal nucleic acids present in a sample containing both fetal and maternal nucleic acids. Fetal fraction is commonly used to characterize cfDNA in maternal blood.

As used herein, the term "chromosome" refers to the genetic carrier of a living cell, derived from a chain of chromatin containing DNA and protein components (especially histones). This article employs the standard, internationally recognized chromosome numbering system for individual human genomes.

As used herein, the term “polynucleotide length” refers to the absolute number of nucleotides in a sequence or region of a reference genome. The term “chromosome length” refers to the known length of a chromosome given in base pairs, for example, the length provided by the NCBI36/hg18 component of the human chromosome available at |genome|.|ucsc|.|edu/cgi-bin/hgTracks?hgsid=167155613&chromInfoPage= on the World Wide Web.

In this paper, the term "subject" refers to both human subjects and non-human subjects, such as mammals, invertebrates, vertebrates, fungi, yeasts, bacteria, and viruses. While the examples in this paper involve humans and the language is primarily human-oriented, the concepts disclosed herein apply to genomes from any plant or animal and can be used in fields such as veterinary medicine, animal science, and research laboratories.

In this article, the term "symptom" refers to "medical symptoms" in a broad sense, which includes all diseases and disorders, as well as injuries and normal health conditions that may affect a person's health, benefit from medical assistance, or influence medical treatment, such as pregnancy.

When used to refer to chromosomal aneuploidy, the term "complete" in this article means the acquisition or loss of the entire chromosome.

When used to refer to chromosomal aneuploidy, the term "part" in this article refers to the acquisition or loss of a portion (i.e., a segment) of a chromosome.

In this article, the term "chimerism" refers to the presence of two cell populations with different chromosomal karyotypes within an individual that has developed from a single fertilized egg. Chimerism can be caused by mutations during development that result in only a subset of adult cells proliferating.

The term "non-chimeric" in this article refers to an organism (such as a human fetus) whose cells are composed of a single chromosomal karyotype.

As used in this article, the term "sensitivity" refers to the probability that a test result will be positive when the conditions of interest are present. Sensitivity can be calculated by dividing the number of true positives by the sum of true positives and false negatives.

As used in this article, the term "specificity" refers to the probability that a test result will be negative when the condition of interest is not present. Specificity can be calculated by dividing the number of true negatives by the sum of true negatives and false positives.

In this article, the term "enrichment" refers to the process of amplifying a polymorphic target nucleic acid contained in a portion of a parent sample and combining the amplified product with the remaining portion of the parent sample after that portion has been removed. For example, the remaining portion of the parent sample may be the original parent sample.

In this article, the term "original maternal sample" refers to a non-enriched biological sample obtained from a pregnant subject (e.g., a woman) from which a portion has been removed to amplify polymorphic target nucleic acids. "Original sample" can be any sample and its processing fraction obtained from a pregnant subject, such as a purified cfDNA sample extracted from a maternal plasma sample.

As used herein, the term "primer" refers to a segregated oligonucleotide that, when placed under conditions that induce the synthesis of the extension product (e.g., conditions including nucleotides, an inducer such as DNA polymerase, and suitable temperature and pH), can act as a starting point for synthesis. Primers are preferably single-stranded molecules that are most efficient in amplification, but double-stranded primers are also acceptable. If double-stranded, the primers are first treated to separate their strands before use in preparing the extension product. Preferably, the primers are oligodeoxyribonucleotides. Primers must be long enough to initiate the synthesis of the extension product in the presence of an inducer. The exact length of the primer will depend on many factors, including temperature, primer source, the purpose of the method, and the parameters used for primer design.

Introduction and Background

CNVs in the human genome significantly influence human susceptibility to disease (Redon et al., Nature, Chapter 23: pp. 444-454 [2006], Shaihio et al., Genome Res, Chapter 19: pp. 1682-1690 [2009]). These diseases include, but are not limited to, cancer, infectious and autoimmune diseases, neurological diseases, metabolic and/or cardiovascular diseases.

CNVs are known to cause genetic disorders through various mechanisms, in most cases leading to gene dosage imbalances or gene damage. In addition to their direct association with genetic disorders, CNVs are known to mediate potentially harmful phenotypic changes. Recently, several studies have reported an increased burden of rare or de novo CNVs in complex diseases such as autism, attention deficit hyperactivity disorder (ADHD), and schizophrenia compared to normal controls, suggesting the potential pathogenicity of rare or unique CNVs (Sebat et al., Chapter 316: pp. 445–449 [2007]; Walsh et al., Chapter 320: pp. 539–543 [2008]). CNVs arise from genomic rearrangements primarily due to deletions, duplications, insertions, and unbalanced translocation events.

Studies have shown that fetal-derived cfDNA fragments are on average shorter than maternal-derived cfDNA fragments. Non-invasive prenatal testing (NIPT) based on NGS data has been successfully implemented. Current methods involve sequencing maternal samples using short reads (25bp-36bp), aligning them to the genome, calculating and normalizing subchromosome coverage, and finally assessing overrepresentation of target chromosomes (13/18/21/X/Y) compared to the expected normalized coverage associated with a normal diploid genome. Therefore, traditional NIPT assays and analyses rely on counting or coverage to assess the likelihood of fetal aneuploidy.

Since maternal plasma samples represent a mixture of maternal and fetal cfDNA, the success of any given NIPT method depends on its sensitivity to detecting copy number variations in samples with low fetal fractions. For counting-based methods, their sensitivity depends on (a) sequencing depth and (b) the ability of data normalization to reduce technical variability. This disclosure provides analytical methods for NIPT and other applications that derive fragment size information from, for example, paired-end reads and use that information in an analytical pipeline. The improved analytical sensitivity provides the ability to apply NIPT methods even with reduced coverage (e.g., reduced sequencing depth), enabling the technique to be used for low-cost detection of average-risk pregnancies.

In some implementations, methods are provided for determining fetal copy number variations (CNVs) using maternal samples containing cell-free maternal and fetal DNA.

This disclosure relates to methods and systems for identifying sample measurements that can reliably make a determination of copy number variation and sample measurements that cannot reliably make such a determination. In other words, the embodiments disclosed in this invention can distinguish between sample measurements that have sufficient information for a credible determination and sample measurements that do not have sufficient information.

In some existing methods for distinguishing between reliable and unreliable sample measurements, the fetal score is used as a metric to determine whether a sample should be excluded. Such methods may run sequencing routines in a specific manner for NIPT samples. The resulting samples have coverage or read count values of non-excluded regions of the genome or chromosome. These methods may also determine a fetal score value. If the determined fetal score value is below a certain threshold, such as 3%, the sample measurement is excluded. At the bottom of such graphs, there is a sample exclusion region that covers sample measurements with a relatively low effective read count and/or a low determined fetal score. In some implementations, the exclusion region includes a step of excluding even relatively high numbers of effective reads for particularly low fetal scores.

As shown in some of the accompanying figures below, the data illustrates various examples of situations where determined fetal score values may be inaccurate for various reasons. The data shows certain trends in fetal score error, such as an increase in error with increasing judgment error (which can be caused by reduced coverage), and an increase in error with decreasing magnitude of the true fetal score value. Because of these problems with using fetal scores as a model for determining which samples to exclude, an alternative technique should be used to determine which samples to exclude.

Fetal fraction influences the determination of fetal cell-free DNA abundance and copy number variation in NIPT. As the fetal fraction decreases in a sample, accurately determining the relative coverage of fetal cell-free DNA becomes more difficult due to signal degradation and relative noise increases with decreasing fetal fraction. Coverage or sequencing depth also has a similar impact on CNV detection. Because both factors jointly affect CNV detection, the minimum fetal fraction required to ensure detection varies with coverage, exhibiting a hyperbolic shape as shown in Figure 1.

Existing methods perform quality control by excluding samples with low coverage. For example, some existing methods set two coverage threshold levels based on fetal score values. In the example shown in Figure 2, the first coverage threshold is set to less than approximately 4.5M non-excluded sites (NES) or valid reads for approximately 5% of fetal scores (line 2002), and the second coverage threshold is set to greater than approximately 2M for approximately 5% of fetal scores (line 2006). A line or curve (2004) connects the two threshold levels.

However, such methods have various limitations. Regarding their impact on CNV detection, they fail to adequately capture the relationship between coverage and fetal score. Furthermore, the thresholds used in these existing methods are static, with the same threshold applied to different samples, populations, and platforms. As shown in Figure 3, different sample populations exhibit different fetal score distributions. Figure 3 illustrates the fetal score distributions for three populations or their samples. The left figure shows the probability density functions for the three populations. The right figure shows the cumulative distribution for the three populations. The cumulative distribution clearly demonstrates that a fixed fetal score threshold of 0.04 excludes different portions of the samples from the three populations.

This disclosure provides methods and systems for quality control (QC) during CNV detection, as shown in Figures 4A, 4B, 15, and 16A. QC can be performed before or after CNV determination to identify samples with fetal fraction levels too low to produce reliable results. Identified samples can be rerun to obtain new reads. If the sequencing depth is increased during rerun, coverage increases, which improves the sample signal or reduces sample noise. The embodiments disclosed herein can account for variations between samples and populations when applied to different sample populations. Furthermore, these embodiments can more effectively identify samples with low fetal fractions and/or low read coverage.

Exemplary Workflow

Unless otherwise specified, the workflows and marked steps in this disclosure may be performed in a different order than those described in the figures and examples. For example, the operation of determining the fetal score shown in box 204 may be performed before or after the operations shown in 202, 206, 208, and 210. Figure 4A illustrates a workflow 200 for determining CNVs using a limit of detection (LOD) QC method according to some specific implementations. In this workflow, a quality check is performed at box 212 to check whether the test sample is within an exclusion region defined by at least a fetal score LOD curve. In many specific implementations, the exclusion region is defined by an LOD curve and a coverage threshold or a fetal score threshold. This step and its downstream steps may be applied to other CNV detection processes described below. This exemplary workflow includes sequencing a test sample containing both maternal and fetal cell-free nucleic acid fragments to obtain sequence reads. See box 202. Various techniques may be used to obtain sequence reads, including but not limited to the sequencing techniques described below.

Then, process 200 proceeds to determine the fetal score of the test sample. In some embodiments, the sequence reads obtained from 202 are used to determine the fetal score. However, other nucleic acids from the same individual may also be used. In some embodiments, various techniques may be used to determine the fetal score of the test sample, including but not limited to those described below. Briefly, in some embodiments, the fetal score of the test sample is determined based on the size of the cell-free nucleic acid fragments. In some embodiments, the fetal score is obtained by obtaining a frequency distribution of cell-free nucleic acid fragment sizes and applying that frequency distribution to a model that correlates the fetal score with the frequency of fragment sizes. In some embodiments, the fetal score is determined based on coverage information of groupings of a reference genome, wherein the reference genome is divided into segments or groups. In some embodiments, the fetal score is calculated by applying coverage values of multiple groups of the reference genome to a model that correlates the fetal score with the coverage of the groups. In some embodiments, groups of overrepresented cell-free fetal nucleic acid fragments are selected from the training samples for alignment. Test reads aligned to the groups are used to determine the fetal score. In some implementations, the fetal score of the test sample is determined based on the coverage information of the sex chromosome grouping (such as the Y chromosome in a maternal sample obtained from a mother carrying a male fetus).

Process 200 also includes receiving sequence reads obtained by sequencing cell-free nucleic acid fragments in the test sample. See box 206. The sequence reads of the cell-free nucleic acid fragments are then aligned with a reference genome including the sequence of interest to provide sequence tags. See box 208.

Procedure 200 continues to determine the coverage of sequence tags for the sequence of interest. See 210. The coverage for each sample is determined. In various embodiments, coverage is determined on all chromosomes (the whole reference genome), on a subset of chromosomes, or at the subchromosome level. Various techniques can be used to determine coverage, including but not limited to those described below. In some embodiments, coverage is determined by dividing the reference genome into multiple groups, determining the number of sequence tags aligned to each group, and using the number of sequence tags in the group of the sequence of interest to determine the coverage of the sequence tags. In some embodiments, the method further includes adjusting the number of sequence tags aligned to groups by taking into account inter-group variation due to factors other than copy number variation. A more detailed description of the methods for determining coverage is provided below.

The next step in process 200 is the QC step. The QC step involves determining whether the test sample is within the exclusion region based on the coverage of the sequence tag determined in 208 and the fetal fraction determined in 204. This exclusion region is defined by a fetal fraction LOD curve. This fetal fraction LOD curve varies with coverage values and indicates the minimum fetal fraction required to achieve the testing criteria for a given coverage level. An example of an LOD curve is shown in Figure 14.

In some implementations, training samples affected by CNVs are used to obtain fetal fraction LOD curves. Affected samples may include computer-simulated samples obtained by combining sequence reads from different individuals. In some implementations, affected samples are obtained by combining physical samples with two or more fetal fractions, thus providing synthetic samples with intermediate fetal fraction values. In some implementations, fetal fraction LOD curves are obtained using both affected and unaffected samples with respect to CNVs. In some implementations, either physical or simulated samples may be used to derive the LOD curve.

In some implementations, the detection criterion is the expected confidence level that the baseline true fetal score is greater than a specified LOD, given an observed fetal score. LOD is a signal (analyte, fetal score, score of indicative condition, etc.) that can define the minimum level of confidence required for detection. In the context of this application, LOD is the lowest level of fetal score (or other analyte) required to define the confidence level for detecting aneuploidy/CNV. It should be noted that two confidence levels are included: the first confidence level is the confidence level of the baseline true fetal score being greater than the specified LOD fetal score, and the second confidence level is the confidence level of the LOD itself. In some implementations, the detection criterion is an X% confidence level that the baseline true fetal score is greater than LOD Y% for the observed fetal score. The baseline true fetal score is an inferred fetal score based on the actual fetal score. In some implementations, X is approximately 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, 99%, or 99.5%. In some implementations, Y represents a detection confidence level of approximately 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, or 99%. In some implementations, X represents 80% and Y represents 95%. In other words, the detection standard is an 80% confidence level that the baseline true fetal score for observing fetal scores is greater than LOD95% (or simply LOD95), which is the smallest fetal score that can detect CNV or aneuploidy 95% of the time. In other implementations, X represents 50% and Y represents 95%.

In some implementations, the designated LOD is determined as the minimum observed fetal score at which Y% of affected samples can be detected. In some implementations, the distribution of the true fetal score (or simulated fetal score) at the observed fetal score coverage is used to obtain the detection criteria for that observed fetal score and that observed coverage. Figure 5 illustrates empirical and hypothetical data illustrating LOD based on statistical concepts. CNV detection is based on the log-likelihood ratio indicating the probability that a sample contains a CNV. The figure above shows the LLR on the Y-axis and the fetal score on the X-axis. Multiple samples are measured at each fetal score level, and the mean and standard deviation are obtained. Given sample data, one can infer the population distribution of the LLR for each level of fetal score. The figure above shows the application of a cutoff value (labeled 502A) to determine CNV. As the fetal score value increases, the LLR score also increases, moving further away from the cutoff value, thus allowing the detection of more samples.

In this example, 2.3% is the lowest fetal fraction at which aneuploidy/CNV can be detected with 95% confidence (one-sided). The observed LLR for affected samples at a fetal fraction of 2.3% is shown as 504A in the upper figure and 504B in the lower figure. The baseline population distribution of the observed data is shown as 506 in the lower figure. The cutoff LLR is shown as 502A in the upper figure and 502B in the lower figure. The lower figure shows that at a fetal fraction of 2.3%, 5% of the baseline population is below the LLR cutoff value, and 95% of the population is above the LLR threshold 502B. Therefore, 95% of the samples in the population were detected as having CNV.

As can be seen in the upper part of Figure 5, the LLR and detection probability increase with the increase of the fetal score. Figure 6 shows the detection probability based on the fetal score. In this example, the detection probability is 95% for a fetal score of 2.3%. Therefore, the LOD95 for the fetal score here is 2.3%.

Returning to Figure 4A, if the test sample is determined not to be within the exclusion region defined by the fetal fractional LOD curve, the method continues to assess the coverage of the sequence of interest to determine the CNV. See the "No" branch from boxes 212 to 214. The procedure ends at this point. If the test sample is determined to be within the LOD exclusion region and it is determined that the sample should be re-sequencing, see the "Yes" branch in boxes 212 and 216. The sample is then re-sequencing, and the procedure from 202 to 212 is repeated. If, at box 216, it is determined that the sequence does not require re-sequencing, such as when it has already been re-sequencing, the procedure ends.

As shown in the experimental data below, the disclosed QC method can help improve both sensitivity and specificity. If one wishes to improve only sensitivity, one can first determine whether the test sample is negative or positive. Only samples determined to be negative can pass the QC process shown in Figure 4B.

Figure 4B illustrates the LOD QC process for CNV detection, which is similar to the process in Figure 4A, except that a decision is performed first before determining whether the sample falls into the exclusion region. See Box 211. If the decision is negative, the process then determines whether the sample is within the exclusion region defined by the LOD curve. See Box 212. If not, the process ends. See the "No" branch of Box 216. If the test sample is within the exclusion region ("Yes" branch of Box 212) and it is determined that resequencing is required, the sample will be resequentialed and the process repeated. If it is determined that the sample does not require resequencing, such as if it has already been resequentialed once, the process ends. See the "No" branch of Box 216.

LOD and LOD curves

Figure 7 illustrates the estimated fetal score, including the error that causes it to deviate from the true fetal score. The solid line shows the distribution of the observed fetal score. The dashed line shows the distribution of the true fetal score. The observed fetal score is inferred from other variables, rather than being measured directly from fetal and maternal DNA fragments. The gray line is the distribution of the fetal score measured directly from the fetal Y chromosome. Because the gray line is directly measured, it is closer to the true fetal score than the observed fetal score.

Figure 8 illustrates the errors in fetal score estimation due to coverage and due to fetal score values. The left figure shows that the standard deviation of the fetal score estimate (a measure of error) decreases as coverage increases. The right figure shows that the standard deviation of the fetal score estimate decreases as the true fetal score increases.

Figure 9 simulates the actual fetal scores at eight levels of observed fetal scores. Figure 9 shows the observed fetal scores at 1% intervals between 0% and 8%. Figure 9 also shows the baseline true distribution of these observed fetal scores. The vertical dashed line 9002 represents the observed fetal score of 0%. The solid line 9004 is the distribution of the actual fetal scores at 0%. The dashed line 9006 indicates the observed fetal score at 8%. The solid line 9008 shows the distribution of the actual fetal scores at 8%. As shown, when the observed fetal scores are low, such as at lines 9002 and 0%, the baseline true fetal score distribution 9004 deviates from the observed fetal scores by a greater distance than the actual fetal score at 8% (9008) deviates from the observed 8% fetal score (9006). Additionally, the peak values in the distribution of lower fetal scores tend to be larger. In other words, as the observed fetal scores increase, the distribution of the actual fetal scores becomes flatter and deviates less from the observed fetal scores.

Figure 10 shows the distribution of the true fetal score relative to the observed fetal score at three different levels of error or coverage (note that coverage is inversely proportional to error). The left plot of Figure 10 is the same as Figure 9. The left plot of Figure 10 simulates a fetal score error of 1%. The middle plot shows the distribution of the fetal score simulated with an error of 1.5%. The right plot shows the distribution of the fetal score simulated with an error of 2%. As can be seen from the figures, the difference between the observed fetal score and the true fetal score is affected by coverage; the larger the error, the greater the difference between the observed fetal score and the true fetal score.

Figure 11 shows the distribution of the observed fetal scores at 2% and their simulated real fetal scores for different errors or coverage levels. The observed fetal scores are indicated by line 1102. The simulated real fetal scores with the minimum error of 1% (or the highest coverage) have distribution 1112. Distribution 1114 is the simulated real fetal scores with 1.5% error or intermediate coverage. Distribution 1116 is the distribution of the simulated real fetal scores with the highest error of 2% or the lowest coverage. The 5th percentile or 95% confidence level of distributions 1112, 1114, and 1116 are marked by lines 1122, 1124, and 1126, respectively. These distributions show that as the error or coverage decreases, the real fetal scores deviate further from the observed fetal scores, and the 95% confidence level also increases. The 20th percentile (1128) and 25th percentile (1130) of distribution 1116 are also shown in Figure 11. They correspond to the 80% and 75% confidence levels, respectively. Similarly, other confidence levels can be determined, such as a 50% confidence level.

Figure 12 shows the distribution of true fetal scores based on the 20th percentile (left) and 25th percentile (right) of the observed fetal score. As the observed fetal score increases, the 20th percentile of the true fetal score also increases accordingly. The left figure shows the 80% confidence data, i.e., the 20th percentile of the true fetal score distribution. The right figure shows the 75% confidence data. In the left figure, the relationship between the 20th percentile true fetal score and the observed fetal score is shown as distinct lines, each indicating a different coverage level (and corresponding FF error level). The left figure shows ten different coverage levels (1M–10M). This figure illustrates the relationship between three variables: the observed fetal score, the true fetal score at the 20th percentile (or 80% confidence level), and coverage. Observing, for example, the left figure, we can see two patterns at the 2% observed FF (corresponding to the FF observed in Figure 11): (a) the 20th percentile (or 80% confidence level) of the true FF is above 2%, and (b) the 20th percentile (or 80% confidence level) of the true FF increases as coverage decreases (or error increases). These two patterns are also visible in Figure 11.

At the lower end of the observed fetal score, the 20th percentile of the true fetal score increases as coverage decreases. This relationship reverses at the higher end of the observed fetal score. To calculate the LOD curve, the LOD95 for a specific coverage is first determined empirically as a value (for the fetal score). This LOD95 value is also the expected (e.g., 20th percentile) value of the true fetal score. Using this value, the observed fetal score can be determined by comparing the true function with the observed function (the line in the left plot of Figure 12).

For example, based on experience, the LOD95 for coverage of 1 million is determined to be 6.50% FF, which is also the 20th percentile of the true fetal score at point 1202. One can determine the observed fetal score as 8% based on the graph (as shown in point 1204). Based on experience, the LOD95 for coverage of 2 million is determined to be 4.59% FF, which is also the 20th percentile of the true fetal score at point 1206. One can determine the observed fetal score as 4.2% based on the graph (as shown in point 1208). One can similarly obtain the observed fetal scores for other coverages from other lines in the graph. These points represent the observed fetal scores required to have a true fetal score exceeding the 80% confidence level of the LOD95 for different coverages.

Figure 13 shows LOD, coverage, and observed fetal score. The table shows the effective read count in the first column from the left, LOD95 in the third column, and the observed fetal score required to achieve an 80% confidence level of a true fetal score higher than LOD95 in the fourth column. The observed fetal score required for a 75% confidence level is shown in the rightmost column, which can be obtained from the data shown in the right panel of Figure 12. The second column shows the FF error for different effective read counts, indicating that the error decreases as coverage increases.

Figure 14 includes two fetal fractional LOD curves that can be obtained from data similar to that shown in Figure 13. In some implementations, the LOD curves are used in conjunction with coverage thresholds, such as the 1M coverage threshold shown in Figure 14.

In some implementations, as shown in Figure 14, the exclusion region lies below the fetal score LOD curve. Numerically, the exclusion region is the area where each point has a lower fetal score or coverage value than the corresponding point on the fetal score LOD curve. In some implementations, the exclusion region is defined by both the fetal score LOD curve and a coverage threshold. The exclusion region lies below both the fetal score LOD curve and the coverage threshold.

In some specific implementations, the methods for generating LOD curves can be summarized as follows: LOD curves are obtained using simulated "samples," where each "sample" includes the following items.

Coverage values sampled from a normal distribution of known coverage values in clinical samples.

• Baseline true FF values determined from the known baseline true FF distribution based on the observed FF distribution in clinical samples.

• The observed FF value is determined by adding the error (corresponding to the coverage value) to the base true FF value.

A large number of such "samples" were simulated. For each observed fetal score (in decimal increments) and at each coverage level (in increments of 1–10M), all corresponding true fetal fractional values in the dataset were collected, and the 100th percentile of these true fetal fractional values was selected minus the X percentile. For each possible combination of observed fetal fractional values and coverage, there is now a single true fetal fractional value to which we can say, given that observed fetal fractional value, that we have at least X% confidence in the true fetal fractional ...

For each coverage level, the FF will be plotted relative to the selected true FF to obtain these plots with a curve for each coverage level.

In addition, LOD studies such as Example 1 describe the relationship between coverage and LOD95 (the derivation of this relationship includes the affected samples). The LOD95(Y) values corresponding to coverage levels 1–10M (in increments) from this plot are then superimposed on the aforementioned plot comparing observed FF to true FF. Each of these horizontal LOD95(Y) lines intersects the corresponding (by coverage) curve on the observed FF vs. true FF plot at a point read from the x-axis to obtain the observed FF required to ensure that the true FF exceeds the X% confidence level of LOD95(Y).

Then, draw the coverage value relative to the observation FF required to obtain the LOD QC curve.

Assess CNV

Methods for determining CNV

Using the sequence coverage values, fragment size parameters, and/or methylation levels provided by the methods disclosed herein, one can determine various genetic conditions related to copy number and CNVs of sequences, chromosomes, or chromosomal segments with improved sensitivity, selectivity, and/or efficiency relative to sequence coverage obtained by conventional methods. For example, in some embodiments, a masked reference sequence is used to determine the presence of any two or more distinct intact fetal chromosomal aneuploidies in a maternal test sample containing fetal and maternal nucleic acid molecules. Exemplary methods provided below align reads to a reference sequence (including a reference genome). Alignment can be performed on unmasked or masked reference sequences, resulting in sequence tags mapped to that reference sequence. In some embodiments, only sequence tags falling on unmasked segments of the reference sequence are considered to determine copy number variations.

In some implementations, assessing the CNV of a nucleic acid sample involves characterizing the state of chromosomal or segmental aneuploidy through one of three types of determinations: “normal” or “unaffected,” “affected,” and “no determination.” Thresholds are typically set for determining normal and affected status. Parameters associated with aneuploidy or other copy number variations are measured in the sample, and the measurements are compared to the thresholds. For replication aneuploidy, an affected determination is made if the chromosomal or segment dose (or other measured sequence content) is higher than a defined threshold set for affected samples. For such aneuploidy, a normal determination is made if the chromosomal or segment dose is lower than a threshold set for normal samples. In contrast, for deletion aneuploidy, an affected determination is made if the chromosomal or segment dose is lower than a defined threshold set for affected samples, and a normal determination is made if the chromosomal or segment dose is higher than a threshold set for normal samples. For example, in the presence of triploidy, a “normal” determination is determined by a parameter value (e.g., test chromosomal dose) below a user-defined reliability threshold, and an “affected” determination is determined by a parameter value (e.g., test chromosomal dose) above a user-defined reliability threshold. A "no-decision" result is determined by a parameter (e.g., the dose of the tested chromosome) that falls between the thresholds for determining "normal" or "affected". The term "no-decision" is used interchangeably with "unclassified".

Parameters that can be used to determine a CNV include, but are not limited to, coverage, fragment size bias/weighted coverage, the fraction or ratio of fragments within a defined size range, and the methylation level of the fragment. As described herein, coverage is obtained from the count of reads aligned to a region of a reference genome, and optionally, this coverage is normalized to produce a sequence tag count. In some embodiments, the sequence tag count may be weighted by fragment size.

In some implementations, the fragment size parameter is biased towards the fragment size characteristic of a genome. The fragment size parameter is a parameter associated with fragment size. The parameter is biased towards fragment size when: 1) the fragment size parameter is advantageously weighted, for example, the count of that size has a greater weight than the count of other sizes; or 2) the parameter is obtained from a value that is advantageously weighted towards fragment size, for example, a ratio obtained from a count that has a greater weight for that size. This size is a characteristic of the genome when it has an enriched or higher concentration of nucleic acids relative to another genome or another part of the same genome.

In some embodiments, a method for determining the presence of any intact fetal chromosomal aneuploidy in a maternal test sample includes (a) obtaining sequence information of fetal and maternal nucleic acids in the maternal test sample; (b) using the sequence information and methods described above to identify the number of sequence tags, sequence coverage metric, fragment size parameter, or another parameter for each chromosome of interest selected from chromosomes 1-22, the X chromosome, and the Y chromosome, and to identify the number of sequence tags or another parameter for one or more normalized chromosome sequences; (c) using the number of sequence tags or the other parameter for each chromosome of interest identified, and the number of sequence tags or the other parameter for each normalized chromosome identified, to calculate a single chromosome dose for each sequence of the chromosome of interest; and (d) comparing each chromosome dose to a threshold, thereby determining whether any intact fetal chromosomal aneuploidy is present in the maternal test sample.

In some embodiments, step (a) above may include sequencing at least a portion of the nucleic acid molecules of the test sample to obtain the sequence information of the fetal and maternal nucleic acid molecules of the test sample. In some embodiments, step (c) includes calculating a single chromosome dose for each chromosome of interest as a ratio of the number of sequence tags or other parameters of each chromosome of interest to the number of sequence tags or other parameters of the identified normalized chromosome sequence. In some other embodiments, the chromosome dose is based on a processed sequence coverage metric or another parameter derived from the number of sequence tags. In some embodiments, the processed sequence coverage metric or another parameter is calculated using only unique, non-redundant sequence tags. In some embodiments, the processed sequence coverage metric is a sequence tag density ratio, which is the number of sequence tags normalized to sequence length. In some embodiments, the processed sequence coverage metric or other parameter is a normalized sequence tag or another normalized parameter, which is the number of sequence tags or other parameters of the sequence of interest divided by the number of sequence tags or other parameters of all or most of the sequence in the genome. In some embodiments, the processed sequence coverage metric or other parameters (such as fragment size parameters) are adjusted according to a global map of the sequence of interest. In some embodiments, the processed sequence coverage metric or other parameters are adjusted based on the intra-sample correlation between the GC content of the test sample and the sequence coverage. In some embodiments, the processed sequence coverage metric or other parameters are generated by a combination of these processes, which are further described elsewhere herein.

In some implementations, chromosome dose is calculated as the ratio of the treated sequence coverage or other parameters of each chromosome of interest to the treated sequence coverage or other parameters of the normalized chromosome sequence.

In any of the above implementation schemes, the complete chromosomal aneuploidy is selected from complete chromosomal triploidy, complete chromosomal monosomy, and complete chromosomal polysomy. The complete chromosomal aneuploidy is selected from the complete aneuploidy of any one of chromosomes 1-22, the X chromosome, and the Y chromosome. For example, the different complete fetal chromosomal aneuploidies are selected from 2-triploid, 8-triploid, 9-triploid, 20-triploid, 21-triploid, 13-triploid, 16-triploid, 18-triploid, 22-triploid, 47,XXX, 47,XYY, and X haploid.

In any of the above embodiments, steps (a) to (d) are repeated for test samples from different maternal subjects, and the method includes determining whether any two or more different intact fetal chromosomal aneuploidies exist in each test sample.

In any of the above embodiments, the method may further include calculating a normalized chromosome value (NCV), wherein the NCV correlates the chromosome dose with the average chromosome dose in the qualified sample set as follows:

Where and are the estimated mean and standard deviation of the dose of the j-th chromosome in the qualified sample set, respectively, and x _{_ij} is the dose of the j-th chromosome in the observed test sample i.

In some implementations, NCV can be calculated "on the fly" as follows: by correlating the chromosome dose of the chromosome of interest in the test sample with the median chromosome dose of the corresponding chromosome in multiplex samples sequenced on the same flow cell.

Where M _{_j} is the estimated median dose of the j-th chromosome in a multiplex sample set sequenced on the same flow cell; is the standard deviation of the dose of the j-th chromosome in one or more multiplex sample sets sequenced on one or more flow cells, and _xi is the observed dose of the j-th chromosome in test sample i. In this embodiment, test sample i is one of the multiplex samples sequenced on the same flow cell where M_{_j} was determined.

In some embodiments, methods are provided for determining the presence of distinct partial fetal chromosomal aneuploidy in a maternal test sample containing fetal and maternal nucleic acids. These methods include procedures similar to those outlined above for detecting complete aneuploidy. However, instead of analyzing complete chromosomes, segments of chromosomes are analyzed. See U.S. Patent Application Publication 2013/0029852, which is incorporated herein by reference.

Figure 15 illustrates methods for determining the presence of copy number variations (CNVs) according to some embodiments. Process 100 shown in Figure 15 uses sequence tag coverage based on the number of sequence tags (i.e., sequence tag counts) to determine CNVs. However, similar to the description above regarding NCV calculation, other variables or parameters (such as size, size ratio, and methylation level) can be used instead of coverage. In some specific embodiments, two or more variables are combined to determine CNVs. Furthermore, coverage and other parameters can be weighted based on the size of the fragment from which the tag originates. For ease of reading, only coverage is mentioned in process 100 shown in Figure 1, but it should be noted that other parameters (such as size, size ratio, methylation level, size-weighted counts, etc.) can be used instead of coverage.

In operations 130 and 135, qualified sequence tag coverage (or a value of another parameter) and test sequence tag coverage (or a value of another parameter) are determined. This disclosure provides a procedure for determining a coverage metric that offers improved sensitivity and selectivity compared to conventional methods. Operations 130 and 135 are marked with an asterisk and bolded in a thick box to indicate that these operations contribute to improvements over the prior art. In some embodiments, the sequence tag coverage metric is normalized, adjusted, trimmed, and otherwise processed to improve the sensitivity and selectivity of the analysis. These procedures are further described elsewhere herein.

In summary, the method utilizes the normalized sequence of qualified training samples to determine the CNV of the test samples. In some embodiments, the qualified training samples are unaffected and have normal copy numbers. The normalized sequence provides a mechanism for normalizing measurements of intra- and inter-run variability. The normalized sequence is identified using sequence information from a qualified sample set obtained from subjects known to contain cells with normal copy numbers of any sequence of interest (such as a chromosome or its segment). Steps 110, 120, 130, 145, and 146 of the implementation of the method shown in Figure 1 outline the determination of the normalized sequence. In some embodiments, the normalized sequence is used to calculate the sequence dose of the test sequence. See step 150. In some embodiments, the normalized sequence is also used to calculate a threshold to which the sequence dose of the test sequence is compared. See step 150. The sequence information obtained from the normalized sequence and the test sequence is used to determine statistically significant chromosomal aneuploidy identification in the test samples (step 160).

Turning to details of methods for determining the presence of copy number variations according to some embodiments, Figure 15 provides a flowchart 100 of an embodiment for determining CNVs (e.g., chromosomes or segments thereof) of a biological sample. In some embodiments, the biological sample is obtained from a subject and contains a mixture of nucleic acids contributed by different genomes. Different genomes may be contributed to the sample by two individuals, for example, by a fetus and a pregnant mother. Alternatively, different genomes may be contributed to the sample by three or more individuals, for example, by two or more fetuses and a pregnant mother. Alternatively, genomes may be contributed to the sample by aneuploid cancer cells and normal euploid cells from the same subject (e.g., plasma samples from cancer patients).

In addition to analyzing patient test samples, one or more normalized chromosomes or one or more normalized chromosomal segments are selected from each possible chromosome of interest. The identification of normalized chromosomes or segments is asynchronous with the normal testing of patient samples, which can be performed in a clinical setting. In other words, normalized chromosomes or segments are identified before testing patient samples. The association between the normalized chromosomes or segments and the chromosomes or segments of interest is stored for use during testing. As explained below, this association is typically maintained over a period of time spanning testing of many samples. The following discussion relates to an implementation scheme for selecting normalized chromosomes or chromosomal segments of interest.

A qualified sample set is obtained to identify qualified normalized sequences and to provide variance values for determining statistically significant CNV identifications in the test samples. In step 110, multiple bioqualified samples are obtained from multiple subjects known to contain cells with normal copy numbers of any one of the sequences of interest. In one embodiment, qualified samples are obtained from a mother carrying a fetus with chromosomes confirmed to have normal copy numbers using cytogenetic methods. Bioqualified samples can be biological fluids (e.g., plasma) or any suitable sample as described below. In some embodiments, the qualified sample contains a mixture of nucleic acid molecules (e.g., cfDNA molecules). In some embodiments, the qualified sample is a maternal plasma sample containing a mixture of fetal and maternal cfDNA molecules. Sequence information of normalized chromosomes and/or their segments is obtained by sequencing at least a portion of the nucleic acids (e.g., fetal and maternal nucleic acids) using any known sequencing method. Preferably, any of the next-generation sequencing (NGS) methods described elsewhere herein are used to sequence fetal and maternal nucleic acids as single or clonal amplified molecules. In various embodiments, qualified samples are processed before and during sequencing as disclosed below. Qualified samples can be processed using devices, systems, and kits as disclosed herein.

In step 120, at least a portion of each of the eligible nucleic acids contained in the eligible sample is sequenced to generate millions of sequence reads (e.g., 36 bp reads) aligned to a reference genome (e.g., hg18). In some embodiments, the sequence reads contain approximately 20 bp, approximately 25 bp, approximately 30 bp, approximately 35 bp, approximately 40 bp, approximately 45 bp, approximately 50 bp, approximately 55 bp, approximately 60 bp, approximately 65 bp, approximately 70 bp, approximately 75 bp, approximately 80 bp, approximately 85 bp, approximately 90 bp, approximately 95 bp, approximately 100 bp, approximately 110 bp, approximately 120 bp, approximately 130 bp, approximately 140 bp, approximately 150 bp, approximately 200 bp, approximately 250 bp, approximately 300 bp, approximately 350 bp, approximately 400 bp, approximately 450 bp, or approximately 500 bp. The anticipated advancement is that technological advancements will enable single-end reads to be greater than 500 bp, thereby allowing reads to be greater than approximately 1000 bp when generating paired-end reads. In one embodiment, the mapped sequence read contains 36 bp. In another embodiment, the mapped sequence read contains 25 bp.

The sequence reads are aligned to a reference genome, and the reads that uniquely map to the reference genome are called sequence tags. Sequence tags that fall on masked regions of the masked reference sequence are not counted for CNV analysis.

In one implementation, at least about 3 × ^10⁶ qualified sequence tags, at least about 5 × 10⁶ qualified sequence tags, at least about 8 × 10⁶ qualified sequence tags, at least about 10 × ^10⁶ qualified sequence tags, at least about 15 × ^10⁶ qualified sequence tags, at least about 20 × ^10⁶ qualified sequence tags, at least about 30 × ^10⁶ qualified sequence tags, at least about 40 × ^10⁶ qualified sequence tags, or at least about ⁵⁰ × ^{10⁶ qualified} sequence tags are obtained from reads uniquely mapped to the reference genome.

In step 130, all tags obtained by sequencing nucleic acids in a qualified sample are counted to obtain qualified sequence tag coverage. Similarly, in step 135, all tags obtained from a test sample are counted to obtain test sequence tag coverage. This disclosure provides a process for determining a coverage metric that offers improved sensitivity and selectivity compared to conventional methods. Steps 130 and 135 are marked with an asterisk and bolded in a thick box to indicate that these steps contribute to improvements over the prior art. In some embodiments, the sequence tag coverage metric is normalized, adjusted, trimmed, and otherwise processed to improve the sensitivity and selectivity of the analysis. These processes are further described elsewhere herein.

When mapping and counting all eligible sequence tags in each eligible sample within the eligible sample, the sequence tag coverage of the sequences of interest (e.g., clinically relevant sequences) in the eligible sample is determined, as well as the sequence tag coverage of other sequences from which normalized sequences are subsequently identified.

In some implementations, the sequence of interest is a chromosome associated with an aneuploidy of the complete chromosome (e.g., chromosome 21), and the eligible normalized sequence is a complete chromosome that is not associated with an aneuploidy and whose sequence tag coverage variation approximates the sequence tag coverage variation of the sequence of interest (i.e., chromosome 21). The selected normalized chromosome can be a single chromosome or a group of chromosomes whose sequence tag coverage variation most closely approximates the sequence of interest. Any one or more chromosomes from chromosomes 1-22, the X chromosome, and the Y chromosome can be the sequence of interest, and one or more chromosomes can be identified as the normalized sequence of any one of chromosomes 1-22, the X chromosome, and the Y chromosome in the eligible sample. As described elsewhere herein, the normalized chromosome can be a single chromosome or a group of chromosomes.

In another implementation, the sequence of interest is a chromosomal segment associated with partial aneuploidy (e.g., chromosomal deletion or insertion, or imbalanced chromosomal translocation), and the normalized sequence is a chromosomal segment (or a group of chromosomal segments) that is not associated with partial aneuploidy and whose sequence tag coverage variation approximates the sequence tag coverage variation of the chromosomal segment associated with partial aneuploidy. The selected normalized chromosomal segment can be one or more chromosomal segments whose sequence tag coverage variation most closely approximates the sequence of interest. Any one or more segments from chromosomes 1-22, the X chromosome, and the Y chromosome can be the sequence of interest.

In other embodiments, the sequence of interest is a chromosomal segment associated with partial aneuploidy, and the normalized sequence is one or more whole chromosomes. In other embodiments, the sequence of interest is a whole chromosome associated with aneuploidy, and the normalized sequence is one or more chromosomal segments unrelated to aneuploidy.

Regardless of whether a single sequence or a group of sequences is identified as a normalized sequence of any one or more sequences of interest in qualified samples, a qualified normalized sequence can be selected to have a sequence tag coverage variation or fragment size parameter that best or effectively approximates the sequence tag coverage variation or fragment size parameter of the sequence of interest determined in qualified samples. For example, a qualified normalized sequence is the sequence that produces the minimum variability among qualified samples when used to normalize the sequence of interest; that is, the variability of the normalized sequence is closest to the variability of the sequence of interest determined in qualified samples. In other words, a qualified normalized sequence is the sequence selected to produce the minimum sequence dose variation (of the sequence of interest) among qualified samples. Therefore, this method selects the sequence that is expected to produce the minimum inter-run chromosome dose variability of the sequence of interest when used as a normalized chromosome.

The normalized sequence of any one or more sequences of interest identified in eligible samples remains the selected normalized sequence used to determine the presence of aneuploidy in test samples over days, weeks, months, and possibly years, provided that the procedures required for generating sequencing libraries and sequencing the samples remain substantially unchanged over time. As described above, the normalized sequence used to determine the presence of aneuploidy is selected based on the variability (or other reasons) of the value of the number of sequence tags or fragment size parameter mapped to samples (e.g., different samples) and sequencing runs (e.g., sequencing runs performed on the same day and/or different days), which best approximates the variability of the sequence of interest used as the normalization parameter. Substantial changes in these procedures will affect the number of tags mapped to all sequences, which in turn will determine which sequence(s) in the same and/or different sequencing runs performed on the same or different days will have the variability closest to that of the sequence of interest across samples, requiring a redetering of the normalized sequence set. Substantial changes to the procedure include changes to the laboratory protocols used to prepare sequencing libraries (including changes related to preparing samples for multiplex sequencing rather than singlex sequencing) and changes to the sequencing platform (including changes to the chemical properties used for sequencing).

In some embodiments, the normalized sequence selected for normalizing a particular sequence of interest is the sequence that best distinguishes one or more qualified samples from one or more affected samples. This means that the normalized sequence is the sequence with the highest distinguishability; that is, the distinguishability of the normalized sequence is such that it provides the best distinction for the sequence of interest in the affected test samples, so as to easily distinguish the affected test samples from other unaffected samples. In other embodiments, the normalized sequence is a sequence with a combination of minimum variability and maximum distinguishability.

The level of distinguishability can be determined as the statistical difference between the sequence dose (e.g., chromosome dose or segment dose) in the qualified sample cohort and the chromosome dose in one or more test samples, as described below and in the examples. For example, distinguishability can be expressed numerically as a t-test value, representing the statistical difference between the chromosome dose in the qualified sample cohort and the chromosome dose in one or more test samples. Similarly, distinguishability can be based on segment dose rather than chromosome dose. Alternatively, distinguishability can be expressed numerically as a normalized chromosome value (NCV), which is the z-score of the chromosome dose, provided that the distribution of the NCV is normal. Similarly, when the chromosome segment is the sequence of interest, the distinguishability of the segment dose can be expressed numerically as a normalized segment value (NSV), which is the z-score of the chromosome segment dose, provided that the distribution of the NSV is normal. When determining the z-score, the mean and standard deviation of the chromosome or segment doses in the qualified sample set can be used. Alternatively, the mean and standard deviation of the chromosome or segment doses in a training set including qualified and affected samples can be used. In other implementations, the normalized sequence is a sequence with the best combination of minimum variability and maximum distinguishability, or small variability and large distinguishability.

This method identifies sequences with inherent similarity characteristics and readily generates similar variations between samples and sequencing runs, and the identified sequences can be used to determine the sequence dose in the test sample.

Determination of sequence dosage

In some implementations, as described in step 146 of Figure 1, one or more chromosome or segment doses of interest are determined among all eligible samples, and a normalized chromosome or segment sequence is identified in step 145. Some normalized sequences are provided before calculating the sequence dose. Then, as further described below, one or more normalized sequences are identified according to various criteria, see step 145. In some implementations, for example, the identified normalized sequence produces minimal sequence dose variability of the sequence of interest among all eligible samples.

In step 146, based on the calculated qualified tag density, the qualified sequence dose (i.e., chromosome dose or segment dose) of the sequence of interest is determined as the ratio of the sequence tag coverage of the sequence of interest to the qualified sequence tag coverage of another sequence from which a normalized sequence is subsequently identified in step 145. The identified normalized sequence is then used to determine the sequence dose in the test sample.

In one implementation, the sequence dose in a qualified sample is a chromosome dose calculated as the ratio of the number of sequence tags or fragment size parameters of the chromosome of interest in the qualified sample to the number of sequence tags of a normalized chromosome sequence. A normalized chromosome sequence can be a single chromosome, a set of chromosomes, a segment of a chromosome, or a set of segments from different chromosomes. Therefore, the chromosome dose of the chromosome of interest in a qualified sample is determined as the ratio of the number of tags on the chromosome of interest to the number of tags on (i) a normalized chromosome sequence consisting of a single chromosome, (ii) a normalized chromosome sequence consisting of two or more chromosomes, (iii) a normalized segment sequence consisting of a single segment of a chromosome, (iv) a normalized segment sequence consisting of two or more segments from a chromosome, or (v) a normalized segment sequence consisting of two or more fragments from two or more chromosomes. Examples of determining the chromosome dose of interest for chromosome 21 based on (i)-(v) are as follows: The chromosome dose of the chromosome of interest (e.g., chromosome 21) is determined as the ratio of sequence tag coverage of chromosome 21 to one of the following sequence tag coverages: (i) each of all remaining chromosomes, i.e., chromosomes 1-20, chromosome 22, chromosome X, and chromosome Y; (ii) all possible combinations of two or more remaining chromosomes; (iii) a segment of another chromosome (e.g., chromosome 9); (iv) two segments of another chromosome, e.g., two segments of chromosome 9; (v) two segments of two different chromosomes, e.g., a segment of chromosome 9 and a segment of chromosome 14.

In another implementation, the sequence dose in the qualified sample is a segmental dose relative to the chromosome dose, calculated as the ratio of the number of sequence tags in the segment of interest (not the entire chromosome) to the number of sequence tags in the normalized segment sequence. The normalized segment sequence can be any one of the normalized chromosome or segment sequences discussed above.

Identification of normalized sequences

In step 145, a normalized sequence of the sequence of interest is identified. In some embodiments, for example, the normalized sequence is a sequence based on a calculated sequence dose, such as one that produces the minimum sequence dose variability of the sequence of interest across all qualified training samples. This method identifies sequences with inherent similarity characteristics and readily produces similar variability between samples and sequencing runs, and the identified sequences can be used to determine the sequence dose in test samples.

Normalized sequences of one or more sequences of interest can be identified in a qualified sample set, and the sequences identified in the qualified samples are then used to calculate the sequence dose of one or more sequences of interest in each test sample in the test samples (step 150) to determine the presence of aneuploidy in each test sample. The identified normalized sequences of the chromosome or region of interest may differ when different sequencing platforms are used and/or when there are differences in the purification of the nucleic acids to be sequenced and/or the preparation of the sequencing libraries. Regardless of the sample preparation and/or sequencing platform used, the use of normalized sequences according to the methods described herein provides a specific and sensitive measure of copy number variation in chromosomes or their regions.

In some implementations, more than one normalized sequence is identified; that is, different normalized sequences of a sequence of interest can be determined, and multiple sequence doses of a sequence of interest can be determined. For example, when using sequence tag coverage of chromosome 14, the chromosomal dose variation of chromosome 21 of interest, such as the coefficient of variation (CV = standard deviation/mean), is minimal. However, two, three, four, five, six, seven, eight, or more normalized sequences can be identified to determine the sequence dose of the sequence of interest in a test sample. For example, chromosomes 7, 9, 11, or 12 can be used as normalized chromosome sequences to determine a second dose of chromosome 21 in any test sample because these chromosomes all have a CV close to that of chromosome 14.

In some implementations, when a single chromosome is selected as the normalized chromosome sequence for the chromosome of interest, the normalized chromosome sequence will be the chromosome that generates the chromosomal dose of the chromosome of interest and has minimal variability across all test samples (e.g., qualified samples). In some cases, the optimal normalized chromosome may not have minimal variability, but may have a qualified dose distribution that best distinguishes one or more test samples from qualified samples; that is, the optimal normalized chromosome may not have minimal variability, but may have maximum distinguishability.

In some embodiments, the normalized sequence includes one or more robust autosomal sequences or fragments thereof. In some embodiments, robust autosomes include all autosomes except the chromosome of interest. In some embodiments, robust autosomes include all autosomes except the X chromosome, Y chromosome, chromosome 13, chromosome 18, and chromosome 21. In some embodiments, robust autosomes include all autosomes except those deviating from a normal diploid state determined from the sample, which can be used to identify cancer genomes with abnormal copy numbers relative to a normal diploid genome.

Determination of aneuploidy in test samples

Based on the identification of normalized sequences in qualified samples, the sequence dose of the sequence of interest in the test sample is determined. The test sample contains a mixture of nucleic acids derived from different genomes of one or more sequences of interest.

In step 115, a test sample is obtained from a subject suspected of or known to carry a clinically relevant CNV of interest. The test sample may be a biological fluid (e.g., plasma) or any suitable sample as described below. As explained, a non-invasive procedure (such as a simple blood draw) may be used to obtain the sample. In some embodiments, the test sample contains a mixture of nucleic acid molecules (e.g., cfDNA molecules). In some embodiments, the test sample is a maternal plasma sample containing a mixture of fetal and maternal cfDNA molecules.

In step 125, at least a portion of the test nucleic acid in the test sample is sequenced as described for qualified samples to generate millions of sequence reads, such as 36 bp reads. In various embodiments, 2 × 36 bp paired end reads are used for paired end sequencing. As described in step 120, the reads generated by sequencing the nucleic acids in the test sample are uniquely mapped to or aligned with a reference genome to generate tags. As described in step 120, at least approximately 3 × 10⁶ qualified sequence tags, at least approximately 5 × ^10⁶ qualified sequence tags, at least approximately 8 × 10⁶ qualified sequence tags, at least approximately 10 × ^10⁶ qualified sequence tags, at least approximately 15 × ^10⁶ qualified sequence tags, at least approximately 20 × ^10⁶ qualified sequence tags, at least approximately 30 × ^10⁶ qualified sequence tags, at least approximately 40 × ^10⁶ qualified sequence tags, or at least approximately ⁵⁰ × ^10⁶ qualified sequence tags are ^obtained from reads uniquely mapped to the reference genome. In some embodiments, the reads generated by the sequencing device are provided in electronic format. Alignment is performed using a computing device as described below. Each read is compared to a reference genome, which is typically very large (millions of base pairs), to identify sites where the read uniquely corresponds to the reference genome. In some embodiments, the alignment procedure allows for limited mismatches between the read and the reference genome. In some cases, mismatches of 1, 2, or 3 base pairs in a read with corresponding base pairs in the reference genome are allowed, but mapping still takes place.

In step 135, the computing device described below is used to count all or most of the tags obtained by sequencing nucleic acids in the test sample to determine the test sequence tag coverage. In some embodiments, each read is aligned to a specific region of the reference genome (in most cases, a chromosome or segment), and the read is converted into a tag by appending site information to the read. As this process unfolds, the computing device can continuously run a count of the number of tags/reads mapped to each region (in most cases, a chromosome or segment) in the reference genome. The counts for each chromosome or segment of interest and each corresponding normalized chromosome or segment are stored.

In some implementations, the reference genome has one or more exclusion regions that are part of the actual biological genome but are not included in the reference genome. Reads that might be aligned to these exclusion regions are not counted. Examples of exclusion regions include regions of long repetitive sequences, similar regions between the X and Y chromosomes, etc. Using the masked reference sequence obtained through the masking techniques described above, only tags on the unmasked segments of the reference sequence are considered for CNV analysis.

In some implementations, when multiple reads are aligned to the same site on a reference genome or sequence, the method determines whether a tag should be counted more than once. It is possible for two tags to have the same sequence and therefore be aligned to the same site on the reference sequence. The method used to count tags may, in some cases, exclude identical tags from the count originating from the same sequencing sample. If a disproportionate number of identical tags are found in a given sample, it indicates a strong bias or other defect in the procedure. Therefore, according to some implementations, the counting method does not count tags from a given sample that are identical to tags from previously counted samples.

Various criteria can be set to select when to ignore identical tags from a single sample. In some embodiments, a defined percentage of the tags to be counted must be unique. If tags exceeding this threshold are not unique, they are ignored. For example, if the defined percentage requires at least 50% uniqueness, identical tags are not counted until the percentage of unique tags in the sample exceeds 50%. In other embodiments, the threshold number of unique tags is at least about 60%. In other embodiments, the threshold percentage of unique tags is at least about 75%, or at least about 90%, or at least about 95%, or at least about 98%, or at least about 99%. The threshold for chromosome 21 can be set to 90%. If 30M tags are compared to chromosome 21, at least 27M of them must be unique. If 3M tags are not unique, and 30 million and one tag is not unique, no counting is performed. Appropriate statistical analysis can be used to select a specific threshold or other criterion for determining when not to count other identical tags. One factor influencing this threshold or other criterion is the relative size of the sequencing sample to the genome size that the tag can be compared to. Other factors include the size of the reading segment and similar considerations.

In one implementation, the number of test sequence tags mapped to the sequence of interest is normalized to the known length of the sequence of interest to which they are mapped, to provide a test sequence tag density ratio. As described for qualified samples, normalization to the known length of the sequence of interest is not required, and this normalization may be included as a step to reduce the number of digits in the numbers to simplify the numbers for human interpretation. When all mapped test sequence tags in the test sample are counted, sequence tag coverage of the sequence of interest (e.g., clinically relevant sequences) in the test sample is determined, as well as sequence tag coverage of additional sequences corresponding to at least one normalized sequence identified in the qualified samples.

In step 150, the test sequence dose of the sequence of interest in the test sample is determined based on the identity of at least one normalized sequence in the qualified sample. In various embodiments, the test sequence dose is calculated using sequence tag coverage of the sequence of interest and its corresponding normalized sequence as described herein. The computing device responsible for this task accesses electronically the association between the sequence of interest and its associated normalized sequence, which may be stored in a database, table, graph, or included as code in program instructions.

As described elsewhere in this document, at least one normalized sequence can be a single sequence or a set of sequences. The sequence dose of the sequence of interest in the test sample is the ratio of the sequence tag coverage of the sequence of interest determined in the test sample to the sequence tag coverage of at least one normalized sequence determined in the test sample, wherein the normalized sequence in the test sample corresponds to the normalized sequence of a specific sequence of interest identified in a qualified sample. For example, if the normalized sequence of chromosome 21 identified in a qualified sample is determined to be a chromosome (e.g., chromosome 14), then the test sequence dose of chromosome 21 (the sequence of interest) is determined as the ratio of the sequence tag coverage of chromosome 21 determined in the test sample to the sequence tag coverage of chromosome 14. Similarly, chromosome doses of chromosomes 13, 18, X, Y, and other chromosomes associated with chromosomal aneuploidy are determined. The normalized sequence of the chromosome of interest can be a single chromosome or a set of chromosomes, or a segment or a set of chromosome segments. As previously mentioned, the sequence of interest can be a portion of a chromosome, such as a chromosome segment. Therefore, the dosage of a chromosome segment can be determined as the ratio of the sequence tag coverage of a segment identified in the test sample to the sequence tag coverage of a normalized chromosome segment identified in the test sample, wherein the normalized segment in the test sample corresponds to a normalized segment (single or a group of segments) of a specific segment of interest identified in a qualified sample. The size of the chromosome segment can range from kilobases (kb) to megabases (Mb) (e.g., about 1 kb to 10 kb, or about 10 kb to 100 kb, or about 100 kb to 1 Mb).

In step 155, a threshold is derived based on the standard deviation of the qualified sequence dose determined in multiple qualified samples and the sequence dose for the sequence of interest in a known aneuploid sample. It should be noted that this operation is typically performed asynchronously with the analysis of patient test samples. This operation may, for example, be performed concurrently with the selection of normalized sequences from qualified samples. Accurate classification depends on the differences between the probability distributions of different categories (i.e., aneuploid types). In some examples, a threshold is selected from an empirical distribution for each type of aneuploidy (e.g., 21-triploidy). As described in the embodiments, possible thresholds are established for classifying 13-triploidy, 18-triploidy, 21-triploidy, and X-haploid aneuploidy; these embodiments describe the use of a method for determining chromosomal aneuploidy by sequencing cfDNA extracted from maternal samples containing a mixture of fetal and maternal nucleic acids. The threshold used to distinguish samples affected by chromosomal aneuploidy may be the same as or different from the thresholds used for different aneuploidy types. As illustrated in the examples, a threshold for each chromosome of interest is determined based on the dose variability of the chromosome of interest in the samples and sequencing runs. The smaller the variability in the chromosome dose for any chromosome of interest, the narrower the dose distribution of the chromosome of interest across all unaffected samples used to set the threshold for determining different aneuploidies.

The process flow associated with classifying patient test samples is returned, and in step 160, the copy number variation of the sequence of interest in the test sample is determined by comparing the test sequence dose of the sequence of interest with at least one threshold established from the qualified sequence dose. This operation can be performed by the same computing device used to measure sequence label coverage and/or calculate segment dose.

In step 160, the calculated dose of the test sequence of interest is compared with a dose set to a threshold selected according to a user-defined "reliability threshold" to classify the sample as "normal," "affected," or "undetermined." "Undetermined" samples are those for which a definitive diagnosis cannot be reliably made. Each type of affected sample (e.g., 21-triploidy, partial 21-triploidy, X-haploidy) has its own threshold, one for classifying normal (unaffected) samples and another for classifying affected samples (although in some cases the two thresholds are the same). As described elsewhere herein, in some cases, if the fetal fraction of nucleic acid in the test sample is sufficiently high, "undetermined" can be converted to "determined" (affected or normal). The classification of the test sequence can be reported by a computing device employed in other operations of this process flow. In some cases, the classification is reported electronically and can be displayed, emailed, texted, or sent to interested parties.

In some implementations, CNV determination includes calculating the NCV or NSV that correlates the chromosomal or segment dose with the average dose of the corresponding chromosomal or segment in a qualified sample set as described above. The CNV can then be determined by comparing the NCV/NSV with a predetermined copy number assessment threshold.

A copy number assessment threshold can be selected to optimize the false positive and false negative rates. A higher copy number assessment threshold reduces the likelihood of false positives. Similarly, a lower threshold reduces the likelihood of false negatives. Therefore, there is a trade-off between a first ideal threshold (above which only true positives are classified) and a second ideal threshold (below which only true negatives are classified).

The threshold setting depends heavily on the chromosomal dose variability of the specific chromosome of interest identified in the unaffected sample set. Variability depends on several factors, including the fraction of fetal cDNA present in the sample. Variability (CV) is determined by the mean or median of the chromosomal dose across the unaffected sample population, along with the standard deviation. Therefore, the threshold used to classify aneuploidy is based on the NCV according to the following formula:

(where and are the estimated mean and standard deviation of the dose of the j-th chromosome in the qualified sample set, respectively, and x _{_ij} is the dose of the j-th chromosome in the observed test sample i.)

The associated fetal scores are as follows:

Therefore, for each NCV of the chromosome of interest, the expected fetal score associated with a given NCV value can be calculated based on the CV of the average and standard deviation of the chromosome ratio of the chromosome of interest in the unaffected sample population.

Subsequently, a decision boundary can be selected based on the relationship between fetal score and NCV value. Samples above this decision boundary are determined to be positive (affected) based on normal distribution quantiles. As described above, in some embodiments, a threshold is set to optimally balance the results of true positive detections and false negative rates. That is, a threshold is selected to maximize the sum of true positives and true negatives, or to minimize the sum of false positives and false negatives.

Some implementations provide a method for prenatal diagnosis of fetal chromosomal aneuploidy in biological samples containing fetal and maternal nucleic acid molecules. The diagnosis is made based on the following process: obtaining sequence information from at least a portion of a mixture of fetal and maternal nucleic acid molecules derived from a biological test sample (e.g., a maternal plasma sample); calculating normalized chromosome doses and/or normalized segment doses of one or more chromosomes of interest and/or segments of interest based on sequencing data; determining statistically significant differences between the chromosome dose of the chromosome of interest and/or the segment dose of the segment of interest in the test sample and thresholds established in multiple qualified (normal) samples, respectively; and providing a prenatal diagnosis based on these statistical differences. A normal or affected diagnosis is made as described in step 160 of the method. In cases where a normal or affected diagnosis cannot be confidently made, a "no decision" is provided.

In some implementations, two thresholds can be selected. A first threshold is selected to minimize the false positive rate; samples above this first threshold are classified as "affected." A second threshold is selected to minimize the false negative rate; samples below this second threshold are classified as "unaffected." Samples with NCVs above the second threshold but below the first threshold can be classified as "suspected aneuploidy" or "undetermined" samples. For suspected aneuploidy or undetermined samples, the presence of aneuploidy can be confirmed independently. The region between the first and second thresholds can be referred to as the "undetermined" region.

In some implementations, the suspected and undetermined thresholds are shown in Table 1. It can be seen that the NCV threshold varies across different chromosomes. In some implementations, as described above, the threshold varies based on the sample's FF. In some implementations, the thresholding techniques applied here help improve sensitivity and selectivity.

Table 1. Threshold Classification of Suspected and Affected NCVs (No Decision Scope)

Suspected Affected chromosome 13 3.5 4.0 chromosome 18 3.5 4.5 chromosome 21 3.5 4.0 X chromosome (XO, XXX) 4.0 4.0 Y chromosome (XX vs. XY) 6.0 6.0

Copy number was determined using the three-pathway process, likelihood ratio, T-statistic, and/or fetal fraction.

Three-pathway process

Figure 16A shows a flowchart of the three-path process for evaluating copy number. The figure includes three overlapping pathways of workflow 700, including pathway 1 (or 713A) analysis of read coverage associated with all sizes of reads, pathway 2 (or 713B) analysis of read coverage associated with shorter reads, and pathway 3 (or 713C) analysis of the relative frequency of shorter reads relative to all reads.

After obtaining the read count, coverage is determined in pathway 713A using reads from all sizes of fragments. Coverage is determined in pathway 713B using reads from short fragments. In pathway 713C, the frequency of reads from short fragments relative to all reads is determined. Relative frequency is also referred to elsewhere herein as size ratio or size fraction. Relative frequency is an example of fragment size characteristics. In some embodiments, a short fragment is a fragment shorter than about 150 base pairs. In various embodiments, a short fragment can be in the size range of about 50 to 150, 80 to 150, or 110 to 150 base pairs. In some embodiments, a third pathway or pathway 713C is optional.

The data from the three pathways 713A, 713B, and 713C are all normalized using operations 714, 716, 718, 719, and 722 to remove variance unrelated to the copy number of the sequence of interest. These normalization operations are boxed in 723. Operation 714 involves normalizing the analytical quantity of the sequence of interest by dividing the analytical quantity by the total value of the quantity of the reference sequence. This normalization step uses values obtained from the experimental samples. Similarly, operations 718 and 722 normalize the analytical quantity using values obtained from the experimental samples. Operations 716 and 719 use values obtained from the unaffected sample training set.

Operation 716 removes the variance of the global wave obtained from the unaffected sample training set. Operation 718 removes the variance of the individual-specific GC variance.

Operation 719 uses Principal Component Analysis (PCA) to remove further variance. The variance removed by PCA is attributed to factors independent of the copy number of the sequence of interest. Analytical quantities (coverage, fragment size ratio, etc.) in each group provide independent variables for the PCA, while samples from the unaffected training set provide values for these independent variables. The training set samples all include samples with the same copy number of the sequence of interest, such as two copies of a somatic chromosome, one copy of the X chromosome (when male samples are used as unaffected samples), or two copies of the X chromosome (when female samples are used as unaffected samples). Therefore, the variance in the samples is not caused by aneuploidy or other copy number variance. PCA on the training set produces principal components independent of the copy number of the sequence of interest. These principal components can then be used to remove variance in the experimental samples independent of the copy number of the sequence of interest.

In some implementations, the variance of one or more principal components is removed from the test sample data using coefficients estimated from data of unaffected samples in regions outside the sequence of interest. In some specific implementations, the region represents all robust chromosomes. For example, PCA is performed on normalized group coverage data of training normal samples to provide principal components corresponding to the dimension of the maximum variance of the data that can be captured. The variance thus captured is independent of copy number variation in the sequence of interest. After obtaining the principal components from the training normal samples, these principal components are applied to the test data. A linear regression model is generated between the groupings from regions outside the sequence of interest, with the test samples as the response variable and the principal components as the dependent variable. The resulting regression coefficients are used to normalize the group coverage of the region of interest by subtracting a linear combination of the principal components defined by the estimated regression coefficients. This removes variance independent of CNV from the sequence of interest. See box 719. The remaining data is used for downstream analysis. Additionally, operation 722 is performed to remove outlier data points.

After the normalization operation in box 723, the coverage values of all groups have been "normalized" to remove sources of variation other than aneuploidy or other copy number variations. In a sense, to detect copy number variations, the grouping of the sequence of interest is enriched or altered relative to other groups. See box 724, which is not an operation but represents the resulting coverage values. The normalization operation in large box 723 can increase the signal and/or reduce noise in the quantity being analyzed. Similarly, the coverage values of short segments of groups have been normalized to remove sources of variation other than aneuploidy or other copy number variations, as shown in box 728, and the relative frequencies (or size ratios) of short segments of groups have been similarly normalized to remove sources of variation other than aneuploidy or other copy number variations, as shown in box 732. Like box 724, boxes 728 and 732 are not operations but represent the coverage and relative frequency values after processing large box 723. It should be understood that the operations in large box 723 can be modified, rearranged, or removed. For example, in some embodiments, PCA operation 719 is not performed. In other embodiments, GC correction operation 718 is not performed. In other embodiments, the order of operations is changed; for example, PCA operation 719 is performed before GC correction operation 718.

The coverage of all segments after normalization and variance removal, as shown in box 724, is used to obtain the t-statistic in box 726. Similarly, the coverage of short segments after normalization and variance removal, as shown in box 728, is used to obtain the t-statistic in box 730, and the relative frequencies of short segments after normalization and variance removal, as shown in box 732, are used to obtain the t-statistic in box 734.

Applying t-statistics to copy number analysis can help improve the accuracy of the analysis. Using only the mean, or both the mean and variance, to distinguish between two distributions does not capture the variance between them. The t-statistic reflects both the mean and variance of the distribution.

In some specific implementations, operation 726 calculates the t-statistic as follows:

Where _x1 is the group coverage of the sequence of interest, _x2 is the group coverage of the reference region/sequence, _s1 is the standard deviation of the coverage of the sequence of interest, _s2 is the standard deviation of the grouping of the reference region, _n1 is the number of groups of the sequence of interest, and _n2 is the number of groups of the reference region.

In some embodiments, the reference region includes all robust chromosomes (e.g., chromosomes other than those most likely to contain aneuploidy). In some embodiments, the reference region includes at least one chromosome other than the sequence of interest. In some simulations, the reference region includes robust chromosomes that do not contain the sequence of interest. In other embodiments, the reference region includes a set of chromosomes that has been determined to provide optimal signal detection capability for the training sample set (e.g., a subset of robust chromosomes). In some embodiments, signal detection capability is based on the ability of the reference region to distinguish groups containing copy number variations from groups that do not contain copy number variations. In some embodiments, the reference region is identified in a manner similar to that used to determine "normalized sequences" or "normalized chromosomes" as described in the section titled "Identification of Normalized Sequences".

Determination of fetal fraction

Returning to Figure 16A, one or more fetal score estimates (box 735) can be combined with any t-statistic in boxes 726, 730, and 734 to obtain likelihood estimates for ploidy cases. See box 736. In some implementations, one or more fetal scores are obtained via any of the procedures 800 in Figure 16B, 900 in Figure 16C, or 1000 in Figure 16D. These procedures can be performed in parallel using the workflow 700 in Figure 16A.

Figure 16B illustrates an exemplary process 800 for determining a fetal score based on coverage information according to some specific embodiments of the present disclosure. Process 800 begins by obtaining coverage information (e.g., sequence dose values) of training samples from a training set. See box 802. Each sample in the training set is obtained from pregnant women known to be carrying male fetuses. That is, the samples contain cfDNA of male fetuses. In some specific embodiments, operation 802 may obtain sequence coverage normalized in a manner different from the sequence doses described herein, or the operation may obtain other coverage values.

Method 800 then includes calculating the fetal score for the training sample. In some implementations, the fetal score can be calculated based on the sequence dose value:

Where _Rxj is the sequence dose of the male sample, and median( _Rxi ) is the median sequence dose of the female sample. In other embodiments, the mean or other measures of central tendency may be used. In some embodiments, FFs may be obtained by other methods, such as the relative frequencies of the X and Y chromosomes. See box 804.

Process 800 also includes dividing the reference sequence into multiple subsequence groups. In some embodiments, the reference sequence is the entire genome. In some embodiments, the groups are 100kb groups. In some embodiments, the genome is divided into approximately 25,000 groups. The process then obtains the coverage of the groups. See box 806. In some embodiments, the coverage used in box 806 is obtained after performing the normalization operation shown in box 723 of Figure 16A. In other embodiments, coverage from different size ranges may be used.

Each group is associated with the coverage of samples in the training set. Therefore, the correlation between the coverage of samples in each group and the fetal score of the sample can be obtained. Process 800 includes obtaining the correlation between the fetal score and the coverage of all groups. See box 808. The process then selects groups with correlation values higher than a threshold. See box 810. In some implementations, groups with the 6000 highest correlation values are selected. The aim is to identify groups that show a high correlation between coverage and fetal score in the training samples. These groups can then be used to predict the fetal score in the test samples. Although the training samples are male samples, the correlation between fetal score and coverage can be applied to both male and female test samples.

Using the selected groupings with high correlation values, the process obtains a linear model that associates fetal score with coverage. See box 812. Each selected grouping provides an independent variable for this linear model. Therefore, the obtained linear model also includes parameters or weights for each group. The weights of these groups are adjusted to fit the model to the data. After obtaining the linear model, process 800 includes applying the coverage data of the test sample to the model to determine the fetal score of the test sample. See box 814. The coverage data of the test sample applied is used for groupings with high correlation between fetal score and coverage.

Figure 16A illustrates a workflow 700 for processing sequence reads, the information of which can be used to obtain fetal score estimates. In some implementations, one or more normalization operations in box 723 are optional. Pathway 1 provides coverage information, which can be used in box 806 of process 800 shown in Figure 16B.

In some implementations, multiple fetal score estimates can be combined to provide a composite fetal score estimate. Various methods can be used to obtain the fetal score estimate. For example, the fetal score can be obtained from coverage information. See process 800 in Figure 16B. In some implementations, the fetal score can also be estimated based on the size distribution of the segments. See process 900 in Figure 16C. In some implementations, the fetal score can also be estimated based on the octamer frequency distribution. See process 1000 in Figure 16D.

In test samples containing cfDNA from male fetuses, fetal scores can also be estimated based on Y chromosome and/or X chromosome coverage. In some embodiments, a composite estimate of the fetal score presuming male fetus is obtained using information selected from: fetal scores obtained from grouped coverage information, fetal scores obtained from fragment size information, fetal scores obtained from Y chromosome coverage, fetal scores obtained from the X chromosome, and any combination thereof. In some embodiments, the presumed sex of the fetus is obtained using Y chromosome coverage. Two or more fetal scores can be combined in various ways to provide a composite estimate of the fetal score. For example, in some embodiments, an average or weighted average method can be used, where the weighting may be based on the statistical confidence level of the fetal score estimate.

In some implementations, a composite estimate of the fetal score for a presumed female fetus is obtained by using information selected from: fetal scores obtained from group coverage information, fetal scores obtained from fragment size information, and any combination thereof.

Figure 16C illustrates a specific implementation of a process for determining fetal scores based on size distribution information. Process 900 begins by obtaining coverage information (e.g., sequence dose values) of male training samples from the training set. See box 902. Process 900 then includes calculating the fetal scores of the training samples using the method described above for box 804. See box 904.

Process 900 continues to divide the size range into multiple groups to provide fragment size-based groupings and determine the read frequencies of these fragment size-based groupings. See box 906. In some embodiments, the frequencies of the fragment size-based groupings are obtained without normalizing the factors shown in box 723. In some embodiments, the frequencies of the fragment size-based groupings are obtained after optionally performing the normalization operation shown in box 723 of Figure 16A. In some embodiments, the size range is divided into 40 groups. In some embodiments, the groupings at the low end include fragments smaller than about 55 base pairs. In some embodiments, the groupings at the low end include fragments in the range of about 50 to 55 base pairs, which excludes information with reads shorter than 50 bp. In some embodiments, the groupings at the high end include fragments larger than about 245 base pairs. In some implementations, the high-terminal groupings include fragments ranging in size from approximately 245 to 250 base pairs, which excludes information with read lengths greater than 250 bp.

Procedure 900 continues by using data from the training samples to obtain a linear model that correlates the fetal score with read frequencies in size-based groups. See box 908. The obtained linear model includes independent variables for read frequencies in size-based groups. The model also includes parameters or weights for each size-based group. The weights of these groups are adjusted to fit the model to the data. After obtaining the linear model, procedure 900 includes applying read frequency data from the test samples to the model to determine the fetal score for the test samples. See box 910.

In some implementations, octamer frequencies can be used to calculate the fetal score. Figure 16D illustrates an exemplary process 1000 for determining a fetal score based on octamer frequency information according to some implementations of this disclosure. Process 1000 begins by obtaining coverage information (e.g., sequence dose values) of male training samples from a training set. See box 1002. Process 1000 then includes calculating the fetal score of the training samples using any of the methods described for box 804. See box 1004.

Process 1000 also includes obtaining the frequencies of octamers (e.g., all possible permutations of 4 nucleotides at 8 positions) from the reads of each training sample. See box 1006. In some embodiments, up to or approximately 65,536 octamers and their frequencies are obtained. In some embodiments, octamer-based frequencies are obtained without normalizing the factors shown in box 723. In some embodiments, octamer frequencies are obtained after optionally performing a normalization operation.

Each octamer is associated with a frequency in the samples of the training set. Therefore, the correlation between the octamer frequency of each octamer and the fetal score of the sample can be obtained. Process 1000 includes obtaining the correlation between the fetal score and the octamer frequency of all octamers. See box 1008. The process then selects octamers with correlation values above a threshold. See box 1010. The aim is to identify octamers that show a high correlation between octamer frequency and fetal score in the training samples. These groupings can then be used to predict the fetal score in the test samples. Although the training samples are male samples, the correlation between fetal score and octamer frequency can be applied to both male and female test samples.

Using selected octamers with high correlation values, the procedure obtains a linear model that correlates fetal score with octamer frequency. See box 1012. Each selected group provides an independent variable for this linear model. Therefore, the obtained linear model also includes parameters or weights for each group. After obtaining the linear model, procedure 1000 includes applying the octamer frequency data of the test sample to the model to determine the fetal score of the test sample. See box 1014.

In some implementations, when calculating the fetal score, the coverage (or other parameters) of different segments that are relevant to different parts of the genome are weighted. Examples of such methods include the SeqFF method described in US Patent Application Publication US2015/0005176 and Kim et al. (2015), Prenatal Diagnosis, Vol. 35, pp. 1–6, the full text of which is incorporated herein by reference for the purpose of calculating the fetal score. In some implementations, groups with higher scores of cell-free nucleic acid fragments are given greater weight in determining the fetal score.

Determine the likelihood ratio

Returning to Figure 16A, in some embodiments, process 700 includes using t-statistics based on all segment coverage provided by operation 726, fetal score estimates provided by operation 726, and t-statistics based on short segment coverage provided by operation 730 to obtain the final ploidy likelihood. These embodiments use a multivariate normal model to combine the results of pathway 1 and pathway 2. In some embodiments used to assess CNV, ploidy likelihood is aneuploidy likelihood, which is the likelihood of a model with an aneuploidy assumption (e.g., triploidy or haploidy) minus the likelihood of a model with an euploidy assumption, wherein the model uses t-statistics based on all segment coverage, fetal score estimates, and t-statistics based on short segment coverage as inputs and provides likelihood as output.

In some specific implementations, ploidy likelihood is expressed as a likelihood ratio. In some specific implementations, the likelihood ratio is modeled as follows:

Where p _₁ represents the likelihood of the data coming from a multivariate normal distribution representing a 3-copy or 1-copy model, p _₀ represents the likelihood of the data coming from a multivariate normal distribution representing a 2-copy model, T _{_short} and T _{_all} are T-scores calculated based on chromosome coverage generated from short fragments and all fragments, and q(ff_{_total} ) is the density distribution of fetal scores (estimated from training data) that takes into account the error associated with the fetal score estimate. The model combines the coverage generated from short fragments with the coverage generated from all fragments, which helps to improve the gap between coverage scores for affected and unaffected samples. In the described implementation, the model also utilizes fetal scores to further improve the ability to distinguish between affected and unaffected samples. Here, the likelihood ratio is calculated using the t-statistic (726) based on all fragment coverage, the t-statistic (730) based on short fragment coverage, and the fetal score estimate provided by procedures 800 (or boxes 726), 900, or 1000 as described above. In some specific implementations, this likelihood ratio is used to analyze chromosomes 13, 18, and 21.

In some specific implementations, the ploidy likelihood obtained by operation 736 uses only the t-statistic based on the relative frequencies of the short segments provided by operation 734 of pathway 3 and the fetal fraction estimates provided by operation 726, procedures 800, 900, or 1000. The likelihood ratio can be calculated using the following formula:

Where p _₁ represents the likelihood of the data coming from a multivariate normal distribution representing a 3-copy or 1-copy model, p _₀ represents the likelihood of the data coming from a multivariate normal distribution representing a 2-copy model, T _{_short-freq} is the T-score calculated based on the relative frequency of the short fragments, and q(ff_{_total}) is the density distribution of the fetal score (estimated from the training data) taking into account the error associated with the fetal score estimate. Here, the likelihood ratio is calculated using the t-statistic (734) based on the relative frequency of the short fragments as described above and the fetal score estimate provided by procedures 800 (or box 726), 900, or 1000. In some specific implementations, this likelihood ratio is used to analyze the X chromosome.

In some implementations, the likelihood ratio is calculated using t-statistics based on coverage of all segments (726), t-statistics based on coverage of short segments (730), and the relative frequency of short segments (734). Alternatively, the likelihood ratio can be calculated by combining the fetal score obtained as described above with the t-statistic. The discriminative power of ploidy assessment can be improved by combining information from any one of the three pathways 713A, 713B, and 713C. In some implementations, different combinations can be used to obtain the chromosome likelihood ratio, such as t-statistics from all three pathways, t-statistics from the first and second pathways, fetal score and three t-statistics, fetal score and one t-statistic, etc. The optimal combination can then be selected based on model performance.

In the various implementations described above, the euploid and aneuploid models use t-statistics as input. However, these models can also use raw or otherwise transformed coverage or abundance values as input and provide likelihood as output. Transformed input values or t-statistics can help improve the model's predictive power, but such transformation is not necessary in all implementations.

In some specific implementations used to assess autosomes, the modeling likelihood ratio represents the likelihood of modeling data obtained from triploid or haploid samples relative to the likelihood of modeling data obtained from diploid samples. In some implementations, this likelihood ratio can be used to determine whether an autosome is triploid or haploid.

In some implementations used to assess sex chromosomes, the likelihood ratios for X haploids and X triploids were evaluated. Additionally, chromosome coverage measures (e.g., CNV or coverage z-score) for the X and Y chromosomes were assessed. In some implementations, decision trees were used to evaluate four values to determine the copy number of the sex chromosomes. In some implementations, decision trees allowed for the identification of ploidy cases of XX, XY, X, XXY, XXX, or XYY.

In some implementations, the likelihood ratio is converted to a log-likelihood ratio, and criteria or thresholds for determining aneuploidy or copy number variation can be empirically set to obtain specific sensitivity and selectivity. For example, when applying the log-likelihood ratio to the training set, it can be set to 1.5 based on the model's sensitivity and selectivity to determine 13-triploidy or 18-triploidy. Furthermore, for example, in some applications, the criterion value for determining triploidy on chromosome 21 can be set to 3.

Samples and Sample Processing

sample

Samples used to determine CNVs (e.g., chromosomal aneuploidy, partial aneuploidy, etc.) may include samples obtained from any cell, tissue, or organ from which copy number variations of one or more sequences of interest are to be determined. Advantageously, the sample contains nucleic acids present in the cells and/or “free” nucleic acids (e.g., cfDNA).

In some implementations, it is advantageous to obtain cell-free nucleic acids, such as cell-free DNA (cfDNA). Free nucleic acids (including free DNA) can be obtained from biological samples using a variety of methods known in the art, including but not limited to plasma, serum, and urine (see, for example, Fan et al., Proc Natl Acad Sci, Chapter 105: pp. 16266-16271 [2008]; Koide et al., Prenatal Diagnosis, Chapter 25: pp. 604-607 [2005]; Chen et al., Nature Med., Chapter 2: pp. 1033-1035 [1996]; Lo et al., Lancet, Chapter 350: pp. 485-487 [1997]; Botezatu et al., Clin Chem., Chapter 46: pp. 1078-1084, 2000; and Su et al., J Mol. Diagn., Chapter 6: pp. 101-107 [2004]). To isolate cell-free DNA from cells in a sample, various methods can be used, including but not limited to fractionation, centrifugation (e.g., density gradient centrifugation), DNA-specific precipitation, or high-throughput cell sorting and/or other separation methods. Commercially available kits for manual and automated isolation of cfDNA are available (Roche Diagnostics, Indianapolis, IN, Qiagen, Valencia, CA, Macherey-Nagel, Duren, DE). Biological samples containing cfDNA have been used in assays to determine the presence of chromosomal abnormalities (e.g., 21-triploidy) by sequencing assays capable of detecting chromosomal aneuploidy and/or multiple polymorphisms.

In various embodiments, the cfDNA present in a sample may be enriched specifically or non-specifically before use (e.g., before preparing a sequencing library). Non-specific enrichment of sample DNA refers to whole-genome amplification of sample genomic DNA fragments that can be used to increase the sample DNA level before preparing a cfDNA sequencing library. Non-specific enrichment can be selective enrichment of one genome present in two genomes in a sample containing more than one genome. For example, non-specific enrichment can be selective for the fetal genome in a maternal sample, which can be obtained by known methods to increase the relative proportion of fetal to maternal DNA in the sample. Alternatively, non-specific enrichment can be non-selective amplification of two genomes present in a sample. For example, non-specific amplification can be amplification of fetal and maternal DNA in a sample containing a mixture of DNA from the fetal and maternal genomes. Methods for whole-genome amplification are known in the art. Degenerate oligonucleotide primer PCR (DOP), primer extension PCR (PEP), and multiple substitution amplification (MDA) are examples of whole-genome amplification methods. In some embodiments, samples containing a mixture of cfDNA from different genomes do not enrich the cfDNA of the genome present in the mixture. In other embodiments, a sample containing a mixture of cfDNA from different genomes is nonspecifically enriched with any one of the genomes present in the sample.

Samples containing the nucleic acids used in the methods described herein typically include biological samples (“test samples”), such as those described above. In some embodiments, the nucleic acids of one or more CNVs to be screened are purified or isolated using any of a variety of well-known methods.

Therefore, in some embodiments, the sample comprises or is composed of purified or isolated polynucleotides, or may include samples such as tissue samples, biofluid samples, cell samples, etc. Suitable biofluid samples include, but are not limited to, blood, plasma, serum, sweat, tears, sputum, urine, phlegm, ear discharge, lymph, saliva, cerebrospinal fluid, lavage fluid, bone marrow suspension, vaginal fluid, cervical lavage fluid, cerebrospinal fluid, ascites, breast milk, respiratory, intestinal and genitourinary secretions, amniotic fluid, and leukocyte-isolated samples. In some embodiments, the sample is readily obtainable through non-invasive procedures, such as blood, plasma, serum, sweat, tears, sputum, urine, phlegm, ear discharge, saliva, or feces. In some embodiments, the sample is a peripheral blood sample or a plasma and/or serum fraction of a peripheral blood sample. In other embodiments, the biological sample is a swab or smear, a biopsy specimen, or a cell culture. In another embodiment, the sample is a mixture of two or more biological samples, such as biological fluid samples, tissue samples, and cell culture samples. As used herein, the terms “blood,” “plasma,” and “serum” explicitly cover their fraction or processing portion. Similarly, in cases where the sample is taken from a biopsy, swab, smear, etc., “sample” explicitly covers the processing fraction or portion derived from a biopsy, swab, smear, etc.

In some implementations, samples may be obtained from sources including, but not limited to, samples from different individuals, samples from different developmental stages of the same or different individuals, samples from different diseased individuals (e.g., individuals with cancer or suspected of having a hereditary disease), normal individuals, samples obtained at different stages of an individual's disease, samples obtained from individuals receiving different treatments for their disease, samples from individuals affected by different environmental factors, samples from individuals susceptible to pathology, and samples obtained from individuals exposed to infectious agents (e.g., HIV).

In one exemplary but non-limiting embodiment, the sample is a maternal sample obtained from a pregnant woman (e.g., a pregnant woman). In this case, the methods described herein can be used to analyze the sample to provide prenatal diagnosis of potential chromosomal abnormalities in the fetus. The maternal sample can be a tissue sample, a biofluid sample, or a cell sample. As a non-limiting example, biofluids include blood, plasma, serum, sweat, tears, sputum, urine, phlegm, ear discharge, lymph, saliva, cerebrospinal fluid, lavage fluid, bone marrow suspension, vaginal fluid, transcervical lavage fluid, cerebrospinal fluid, ascites, breast milk, respiratory, intestinal, and genitourinary secretions, and leukocyte separation samples.

In another exemplary but non-limiting embodiment, the parent sample is a mixture of two or more biological samples, such as biological fluid samples, tissue samples, and cell culture samples. In some embodiments, the sample is a sample readily obtainable by non-invasive surgery, such as blood, plasma, serum, sweat, tears, sputum, urine, breast milk, phlegm, earwax, saliva, and feces. In some embodiments, the biological sample is a peripheral blood sample and/or its plasma and serum components. In other embodiments, the biological sample is a swab or smear, a biopsy specimen, or a cell culture sample. As disclosed above, the terms “blood,” “plasma,” and “serum” explicitly cover their fractions or processing portions. Similarly, in cases where the sample is taken from a biopsy, swab, smear, etc., “sample” explicitly covers the processing fraction or portion derived from a biopsy, swab, smear, etc.

In some embodiments, samples may also be obtained from in vitro cultured tissues, cells, or other polynucleotide-containing sources. Cultured samples may be obtained from sources including, but not limited to, cultures (e.g., tissues or cells) maintained in different media and conditions (e.g., pH, pressure, or temperature), cultures (e.g., tissues or cells) maintained for different durations, cultures (e.g., tissues or cells) treated with different factors or reagents (e.g., candidate drugs or modulators), or cultures of different types of tissues and/or cells.

Methods for isolating nucleic acids from biological sources are well known and will vary depending on the nature of the source. Those skilled in the art can readily isolate nucleic acids from the sources desired by the methods described herein. In some cases, it may be advantageous to fragment the nucleic acid molecules in the nucleic acid sample. Fragmentation can be random or it can be specific, such as by digestion with restriction endonucleases. Methods for random fragmentation are well known in the art and include, for example, restriction DNase digestion, alkaline treatment, and physical shearing. In one embodiment, sample nucleic acids are obtained from unfragmented cfDNA.

Sequencing library preparation

In one embodiment, the methods described herein may utilize next-generation sequencing (NGS) technology, which allows for the individual sequencing of multiple samples in a single sequencing run, either as genomic molecules (i.e., singlet sequencing) or as a pooled sample containing index genomic molecules (e.g., multiplex sequencing). These methods can generate up to billions of DNA sequence reads. In various embodiments, next-generation sequencing (NGS) technologies, such as those described herein, may be used to determine the sequences of genomic nucleic acids and/or index genomic nucleic acids. In various embodiments, one or more processors, as described herein, may be used to perform analysis of large volumes of sequence data obtained using NGS.

In various implementation schemes, the use of such sequencing technologies does not include the preparation of sequencing libraries.

However, in some embodiments, the sequencing methods envisioned herein include the preparation of sequencing libraries. In one exemplary method, sequencing library preparation involves randomly acquiring adaptor-modified DNA fragments (e.g., polynucleotides) to be sequenced. Polynucleotide sequencing libraries can be prepared from DNA or RNA (including equivalents or analogs of DNA or cDNA, such as complementary or copy DNA or cDNA generated from an RNA template) by the action of reverse transcriptase. Polynucleotides may be initiated in double-stranded form (e.g., dsDNA, such as genomic DNA fragments, cDNA, PCR amplification products, etc.), or in some embodiments, polynucleotides may be initiated in single-stranded form (e.g., ssDNA, RNA, etc.) and have been converted to dsDNA form. For example, in some embodiments, single-stranded mRNA molecules may be copied into double-stranded cDNA suitable for preparing sequencing libraries. The precise sequence of the primary polynucleotide molecule is generally not important to the library preparation method and may be known or unknown. In one embodiment, the polynucleotide molecule is a DNA molecule. More specifically, in some embodiments, the polynucleotide molecule represents the entire complementary genetic sequence of an organism or substantially the entire complementary genetic sequence of an organism, and is a genomic DNA molecule (e.g., cellular DNA, cell-free DNA (cfDNA), etc.), which typically includes intron sequences and exon sequences (coding sequences), as well as non-coding regulatory sequences such as promoter and enhancer sequences. In some embodiments, the primary polynucleotide molecule includes human genomic DNA molecules, such as cfDNA molecules present in the peripheral blood of pregnant subjects.

The preparation of sequencing libraries on some NGS sequencing platforms is facilitated by using polynucleotides that encompass a specific range of fragment sizes. The preparation of such libraries typically involves fragmenting large polynucleotides (e.g., cellular genomic DNA) to obtain polynucleotides within the desired size range.

Fragmentation can be achieved by any of a variety of methods known to those skilled in the art. For example, fragmentation can be achieved by mechanical methods, including but not limited to atomization, sonication, and water shearing. However, mechanical fragmentation typically involves cutting the DNA backbone at C-O, P-O, and C-C bonds, resulting in a heterogeneous mixture with blunt ends and 3'- and 5'-protruding ends of broken C-O, P-O, and/or C-C bonds (see, for example, Alnemri and Liwack, J Biol. Chem, Chapter 265: pp. 17323-17333 [1990]; Richards and Boyer, J Mol Biol, Chapter 11: pp. 327-240 [1965]). This heterogeneous mixture may require repair because it may lack the 5'-phosphate necessary for subsequent enzymatic reactions (e.g., the ligation of sequencing adaptors required for the preparation of DNA for sequencing).

In contrast, cfDNA typically exists in fragments of less than about 300 base pairs, so fragmentation is usually not necessary for generating sequencing libraries from cfDNA samples.

Typically, whether polynucleotides are forcibly fragmented (e.g., in vitro fragmentation) or exist naturally as fragments, they are converted into blunt-ended DNA with 5'-phosphate and 3'-hydroxyl groups. Standard protocols, such as those using Illumina platforms described elsewhere herein, instruct the user to perform end repair on the sample DNA, purify the end-repaired product before dA-tailing, and purify the dA-tailed product before the adaptor ligation step in library preparation.

Various embodiments of the sequencing library preparation methods described herein do not require the performance of one or more steps typically required by standard protocols to obtain modified DNA products that can be sequenced by NGS. Examples of simplified methods (ABB methods), one-step methods, and two-step methods for preparing sequencing libraries can be found in patent application 13/555,037, filed July 20, 2012, the entire contents of which are incorporated herein by reference.

Labeled nucleic acids used to track and verify sample integrity

In various implementations, sample integrity verification and sample tracking can be achieved, for example, by sequencing a mixture of sample genomic nucleic acids (e.g., cfDNA) and accompanying labeled nucleic acids introduced into the sample prior to processing.

Labeled nucleic acids can be combined with a test sample (e.g., a biological sample) and processed including one or more of the following steps: fractionation of the biological sample (e.g., obtaining a substantially free plasma fraction from a whole blood sample), purification of nucleic acids from the fractionated (e.g., plasma) or unfractionated biological sample (e.g., tissue sample), and sequencing. In some embodiments, sequencing includes preparing a sequencing library. The sequence or combination of sequences of the labeling molecules combined with the source sample is selected to be unique to that source sample. In some embodiments, the unique labeling molecules in the sample all have the same sequence. In other embodiments, the unique labeling molecules in the sample are multiple sequences, such as combinations of two, three, four, five, six, seven, eight, nine, ten, fifteen, twenty, or more different sequences.

In one implementation, multiple labeled nucleic acid molecules having the same sequence can be used to verify the integrity of a sample. Alternatively, multiple labeled nucleic acid molecules can be used to verify the identity of a sample, wherein the labeled nucleic acid molecules have at least two, three, four, five, six, seven, eight, nine, ten, eleven, twelve, thirteen, fourteen, fifteen, sixteen, seventeen, eighteen, nineteen, eleven, thirteen, eleventeen, thirteenth ... or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteenth, or thirteen Alternatively, a first sample may be labeled with a marker nucleic acid molecule that all have sequence A, and a second sample may be labeled with a mixture of sequences B and C, wherein sequences A, B, and C are marker molecules with different sequences.

Labeled nucleic acids can be added to a sample at any stage of sample preparation, prior to library preparation (if a library is to be prepared) and sequencing. In one embodiment, the labeling molecule can be combined with an untreated source sample. For example, the labeling nucleic acid can be provided in a collection tube used to collect a blood sample. Alternatively, the labeling nucleic acid can be added to a blood sample after blood collection. In one embodiment, the labeling nucleic acid is added to a container used to collect a biological fluid sample, such as adding the labeling nucleic acid to a blood collection tube used to collect a blood sample. In another embodiment, the labeling nucleic acid is added to a fraction of a biological fluid sample. For example, the labeling nucleic acid is added to the plasma and/or serous components of a blood sample (e.g., a maternal plasma sample). In another embodiment, the labeling molecule is added to a purified sample, such as a nucleic acid sample purified from a biological sample. For example, the labeling nucleic acid is added to purified maternal and fetal cfDNA samples. Similarly, the labeling nucleic acid can be added to a biopsy specimen prior to specimen processing. In some embodiments, the labeling nucleic acid can be combined with a carrier for delivering the labeling molecule to cells of a biological sample. Cell delivery carriers include pH-sensitive and cationic liposomes.

In various embodiments, the marker molecule has anti-genomic sequences that are not present in the genome of the biologically derived sample. In one exemplary embodiment, the marker molecule used to verify the integrity of a human biologically derived sample has a sequence not present in the human genome. In another embodiment, the marker molecule has a sequence not present in the source sample and any one or more other known genomes. For example, the marker molecule used to verify the integrity of a human biologically derived sample has a sequence not present in both the human genome and the mouse genome. Alternatives allow for the verification of the integrity of test samples containing two or more genomes. For example, the integrity of a human cell-free DNA sample obtained from a subject affected by a pathogen (e.g., bacteria) can be verified using a marker molecule having a sequence not present in both the human genome and the genome of the affected bacteria. Genomic sequences of a variety of pathogens (e.g., bacteria, viruses, yeasts, fungi, protozoa, etc.) are publicly available at ncbi.nlm.nih.gov/genomes on the World Wide Web. In another embodiment, the marker molecule is a nucleic acid having a sequence not present in any known genome. The sequence of the marker molecule can be randomly generated by an algorithm.

In various implementations, the marker molecule can be naturally occurring deoxyribonucleic acid (DNA), ribonucleic acid, or artificial nucleic acid analogs (nucleic acid mimics). Artificial nucleic acid analogs include peptide nucleic acids (PNA), morpholinonucleotides, locked nucleic acids, diol nucleic acids, and threonic acid, which differ from naturally occurring DNA or RNA in that they lack a phosphodiester backbone or have altered DNA mimic backbones. Deoxyribonucleic acid can be derived from naturally occurring genomes or generated in the laboratory using enzymes or through solid-phase chemical synthesis. Chemical methods can also be used to generate DNA mimics that do not exist in nature. Available DNA derivatives in which the phosphodiester bonds have been replaced, but the deoxyribose has been retained, include, but are not limited to, DNA mimics having a backbone formed by thioformyl acetals or carboxamide bonds, which have shown to be good structural DNA mimics. Other DNA mimics include morpholino derivatives and peptide nucleic acids (PNA), which contain a pseudopeptide backbone based on N-(2-aminoethyl)glycine (Ann Rev Biophys Biomol Struct, Chapter 24: pp. 167-183 [1995]). PNA is an excellent structural mimic of DNA (or ribonucleic acid [RNA]), and PNA oligomers can form very stable double-stranded structures with Watson-Crick complementary DNA and RNA (or PNA) oligomers, and they can also bind to targets in double-stranded DNA via helical invasion (Mol Biotechnol, Chapter 26: pp. 233-248 [2004]). Another good structural mimic/analyst for DNA can be used as a marker molecule is phosphate-thioester DNA, in which one of the oxygen atoms in the non-bridging oxygen is replaced by sulfur. This modification reduces the activity of endonucleases and exonucleases 2 (including 5' to 3' and 3' to 5' DNAPOL 1 exonucleases), nucleases S1 and P1, RNases, serum nucleases, and snake venom phosphodiesterase.

The length of the marker molecule may be different from or similar to the length of the sample nucleic acid; that is, the length of the marker molecule may be similar to the length of the sample genomic molecule, or it may be greater than or less than the length of the sample genomic molecule. The length of the marker molecule is measured by the number of nucleotide or nucleotide analog bases constituting the marker molecule. Separation methods known in the art can be used to distinguish marker molecules with a length different from that of the sample genomic molecule from the source nucleic acid. For example, the length difference between the marker and sample nucleic acid molecules can be determined by electrophoretic separation (e.g., capillary electrophoresis). Size differences can be beneficial for the quantification and evaluation of the quality of the marker and sample nucleic acid. Preferably, the marker nucleic acid is shorter than the genomic nucleic acid and has sufficient length to exclude them from mapping to the sample genome. For example, in cases where a 30-base human sequence is required to uniquely map to the human genome. Therefore, in some embodiments, the length of the marker molecule used for sequencing bioassays of human samples should be at least 30 bp.

The selection of marker molecule length is primarily determined by the sequencing technology used to verify the integrity of the source sample. The length of the genomic nucleic acid in the sequenced sample may also be considered. For example, some sequencing technologies employ the clonal amplification of polynucleotides, which may require the genomic polynucleotides to be cloned to have a minimum length. For instance, sequencing using an Illumina GAII sequencer involves the in vitro clonal amplification of a polynucleotide with a minimum length of 110 bp via bridge PCR (also known as cluster amplification), ligating an adaptor to the polynucleotide to provide nucleic acid of at least 200 bp and less than 600 bp for cloning and sequencing. In some embodiments, the length of the marker molecule ligated to the adaptor is between about 200 bp and about 600 bp, about 250 bp and about 550 bp, about 300 bp and about 500 bp, or about 350 bp and about 450 bp. In other embodiments, the length of the marker molecule ligated to the adaptor is about 200 bp. For example, when sequencing fetal cfDNA present in maternal samples, the length of the marker molecule may be selected to be similar to the length of the fetal cfDNA molecule. Therefore, in one embodiment, the length of the marker molecule used in the assay can be about 150 bp, about 160 bp, 170 bp, about 180 bp, about 190 bp, or about 200 bp. This assay involves massively parallel sequencing of cfDNA in a maternal sample to determine the presence of fetal chromosomal aneuploidy; preferably, the marker molecule is about 170 pp. Other sequencing methods (e.g., SOLiD sequencing, Polony sequencing, and 454 sequencing) use emulsion PCR to clone and amplify DNA molecules for sequencing, and each technique specifies minimum and maximum lengths of the molecules to be amplified. The length of the marker molecule to be sequenced as a cloned and amplified nucleic acid can be up to about 600 bp. In some embodiments, the length of the marker molecule to be sequenced can be greater than 600 bp.

Single-molecule sequencing technology does not employ clonal amplification of molecules and can sequence nucleic acids over a very wide range of template lengths, in most cases without requiring the molecule to be sequenced to have any specific length. However, the sequence yield per unit mass depends on the number of 3' hydroxyl groups, so a relatively short sequencing template is more efficient than a long one. If starting with nucleic acids longer than 1000 nt, it is generally recommended to cleave the nucleic acids to an average length of 100 to 200 nt to generate more sequence information from the same mass of nucleic acids. Therefore, the length of labeled molecules can range from tens to thousands of bases. The length of labeled molecules used for single-molecule sequencing can be up to approximately 25 bp, up to approximately 50 bp, up to approximately 75 bp, up to approximately 100 bp, up to approximately 200 bp, up to approximately 300 bp, up to approximately 400 bp, up to approximately 500 bp, up to approximately 600 bp, up to approximately 700 bp, up to approximately 800 bp, up to approximately 900 bp, up to approximately 1000 bp, or longer.

The length of the selected marker molecule was also determined by the length of the sequenced genomic nucleic acid. For example, cfDNA circulates in human bloodstream as a fragment of cellular genomic DNA. Fetal cfDNA molecules present in maternal plasma are typically shorter than maternal cfDNA molecules (Chan et al., Clin Chem, Chapter 50: p. 8892 [2004]). Size fractionation of circulating fetal DNA has confirmed that the average length of circulating fetal DNA fragments is <300 bp, while maternal DNA is estimated to be between approximately 0.5 Kb and 1 Kb (Li et al., Clin Chem, Chapter 50: pp. 1002-1011 [2004]). These findings are consistent with those of Fan et al., who used NGS to determine that fetal cfDNA is rarely >340 bp (Fan et al., Clin Chem, Chapter 56: pp. 1279-1286 [2010]). DNA isolated from urine using standard silica-based methods consists of two fractions: high-molecular-weight DNA derived from exfoliated cells and low-molecular-weight (150-250 base pairs) fractions of renal DNA (Tr-DNA) (Botezatu et al., Clin Chem. Chapter 46: pp. 1078-1084, 2000; and Su et al., J Mol. Diagn. Chapter 6: pp. 101-107, 2004). Application of newly developed techniques for isolating cell-free nucleic acids from bodily fluids to the isolation of renal nucleic acids has shown that DNA and RNA fragments present in urine are much shorter than 150 base pairs (US Patent Application Publication 20080139801). In embodiments, where cfDNA is sequenced genomic nucleic acid, the selected marker molecule can be at most approximately the length of cfDNA. For example, the length of the marker molecule used in a maternal cfDNA sample to be sequenced as a single nucleic acid molecule or cloned amplified nucleic acid can be between approximately 100 bp and 600 bp. In other embodiments, the sample genomic nucleic acid is a fragment of a larger molecule. For example, the sequenced sample genomic nucleic acid is fragmented cellular DNA. In embodiments, when sequencing fragmented cellular DNA, the length of the marker molecule may be at most the length of the DNA fragment. In some embodiments, the length of the marker molecule is the minimum length required to at least uniquely map the sequence read to the appropriate reference genome. In other embodiments, the length of the marker molecule is the minimum length required to exclude the marker molecule from mapping to the sample reference genome.

In addition, marker molecules can be used to verify samples that have not been determined by nucleic acid sequencing and can be verified by common biotechnologies other than sequencing, such as real-time PCR.

Sample controls (e.g., positive controls used during sequencing and/or analysis).

In various implementations, such as those described above, the labeled sequence introduced into the sample can be used as a positive control to verify the accuracy and validity of sequencing and subsequent processing and analysis.

Therefore, compositions and methods are provided for providing a positive control (IPC) during the sequencing of DNA in a sample. In some embodiments, a positive control is provided for sequencing cfDNA in a sample containing a mixture of genomes. The IPC can be used to correlate baseline offsets of sequence information obtained from sets of different samples (e.g., samples sequenced at different times on different sequencing runs). Thus, for example, the IPC can correlate sequence information obtained from a maternal test sample with sequence information obtained from a set of qualified samples sequenced at different times.

Similarly, in the case of segment analysis, IPC can correlate sequence information of a specific segment obtained from a subject with sequences obtained from a qualified sample set (of similar sequences) sequenced at different times. In some embodiments, IPC can correlate sequence information of a specific cancer-related locus obtained from a subject with sequence information obtained from a qualified sample set (e.g., from known amplifications/deletions, etc.).

In addition, IPCs can be used as markers to track samples during sequencing. IPCs can also provide qualitative positive sequence dose values (e.g., NCVs) for one or more aneuploidies of the chromosome of interest (e.g., 21-triploid, 13-triploid, 18-triploid) to provide appropriate interpretation and ensure the reliability and accuracy of the data. In some embodiments, IPCs containing nucleic acids from both the male and female genomes can be generated to provide dose to the X and Y chromosomes in maternal samples.

The type and number of in-process controls depend on the type or nature of the test required. For example, for a test that requires sequencing of DNA from a sample containing a mixture of genomes to determine the presence of chromosomal aneuploidy, in-process controls may include DNA obtained from a sample known to contain the same chromosomal aneuploidy being tested. In some embodiments, IPCs contain DNA from samples known to contain aneuploidy of the chromosome of interest. For example, an IPC for a test to determine the presence of fetal triploidy (e.g., 21-triploidy) in a maternal sample contains DNA obtained from an individual with 21-triploidy. In some embodiments, IPCs contain a mixture of DNA obtained from two or more individuals with different aneuploidies. For example, an IPC for a test to determine the presence of 13-triploidy, 18-triploidy, 21-triploidy, and X haploidy contains a combination of DNA samples obtained from each pregnant woman carrying a fetus and one of the triploids being tested. In addition to complete chromosomal aneuploidy, IPCs may also be generated to provide a positive control for a test used to determine the presence of partial aneuploidy.

An IPC (Integrated Biomarker) can be generated using a mixture of cellular genomic DNA obtained from two subjects, one of whom is a contributor to the aneuploid genome. For example, an IPC can be generated by combining genomic DNA from a male or female subject carrying a trisomy chromosome with genomic DNA from a female subject known not to carry a trisomy chromosome; this IPC serves as a control for a test to determine fetal triploidy (21-trisomy). Genomic DNA can be extracted from the cells of both subjects and cut to provide fragments between approximately 100 bp and 400 bp, approximately 150 bp and 350 bp, or approximately 200 bp and 300 bp to mimic circulating cfDNA fragments in maternal samples. The proportion of fragmented DNA from subjects carrying aneuploidy (e.g., trisomy 21) is selected to mimic the proportion of circulating fetal cfDNA present in maternal samples to provide an IPC containing a mixture of fragmented DNA comprising approximately 5%, approximately 10%, approximately 15%, approximately 20%, approximately 25%, and approximately 30% DNA from subjects carrying aneuploidy. The IPC may contain DNA from different subjects each carrying different aneuploidies. For example, the IPC may contain approximately 80% unaffected female DNA, and the remaining 20% may be DNA from three different subjects each carrying trisomy 21, trisomy 13, and trisomy 18. The mixture of fragmented DNA is prepared for sequencing. Processing of the mixture of fragmented DNA may include preparing a sequencing library that can be sequenced in singleton or multiplex mode using any massively parallel method. Stock solutions of genomic IPCs can be stored and used in multiple diagnostic tests.

Alternatively, IPCs can be generated using cfDNA obtained from a mother known to be carrying a fetus with a known chromosomal aneuploidy. For example, cfDNA can be obtained from a pregnant woman carrying a fetus with a 21-triploidy. The cfDNA is extracted from the maternal sample, cloned into a bacterial vector, and grown in bacteria to provide a continuous source of IPCs. The DNA can be extracted from the bacterial vector using restriction enzymes. Alternatively, the cloned cfDNA can be amplified, for example, by PCR. The IPC DNA can be processed in the same run as the cfDNA from the test sample for sequencing, which will be analyzed for the presence of chromosomal aneuploidy.

While the generation of IPCs has been described above in the context of triploidy, it should be understood that IPCs can be generated to reflect other partial aneuploidy, including, for example, various segment amplifications and/or deletions. Thus, for example, in cases where various cancers are known to be associated with a particular amplification (e.g., breast cancer associated with 20Q13), IPCs incorporating these known amplifications can be generated.

sequencing methods

As described above, the prepared sample (e.g., a sequencing library) is sequenced as part of a procedure for identifying copy number variations. Any of a variety of sequencing technologies can be used.

Some sequencing technologies are commercially available, such as the hybridization-while-sequencing platform from Affymetrix (Sunnyvale, CA), the synthesis-while-sequencing platform from 454 Life Sciences (Bradford, CT), Illumina/Solexa (Hayward, CA), and Helicos Biosciences (Cambridge, MA), and the ligation-while-sequencing platform from Applied Biosystems (Foster City, CA), as described below. In addition to single-molecule sequencing using Helicos Biosciences' synthesis-while-sequencing, other single-molecule sequencing technologies include, but are not limited to, Pacific Biosciences' SMRT ^™ technology, ION TORRENT ^™ technology, and nanopore sequencing, for example, developed by Oxford Nanopore Technologies.

While the automated Sanger method is considered a "first-generation" technology, Sanger sequencing, including automated Sanger sequencing, can also be used in the methods described herein. Other suitable sequencing methods include, but are not limited to, nucleic acid imaging techniques such as atomic force microscopy (AFM) or transmission electron microscopy (TEM). Exemplary sequencing techniques are described in more detail below.

In one exemplary but non-limiting embodiment, the method described herein includes using Illumina's sequencing-by-synthesis and sequencing chemistry based on reversible terminators to obtain sequence information of nucleic acids in a test sample, such as cfDNA from a maternal sample, cfDNA from a subject screening for cancer, or cellular DNA (e.g., as described in Bentley et al., Nature, Chapter 6: pp. 53-59 [2009]). The template DNA can be genomic DNA, such as cellular DNA or cfDNA. In some embodiments, genomic DNA from isolated cells is used as a template and fragmented into lengths of several hundred base pairs. In other embodiments, cfDNA is used as a template and does not need to be fragmented because cfDNA exists as a short fragment. For example, fetal cfDNA circulates in the bloodstream in fragments approximately 170 base pairs (bp) in length (Fan et al., Clin Chem, Chapter 56: pp. 1279-1286 [2010]), and this DNA does not need to be fragmented prior to sequencing. Illumina's sequencing technology relies on ligating fragmented genomic DNA to an optically transparent surface on which the anchor oligonucleotide is bound. The template DNA is end-repaired to produce 5'-phosphorylated blunt ends, and a single A base is added to the 3' end of the blunt-ended phosphorylated DNA fragment using the polymerase activity of the Klenow fragment. This addition prepares the DNA fragment for ligation to an oligonucleotide adaptor, which has a single T base overhang at its 3' end to improve ligation efficiency. The adaptor oligonucleotide is complementary to the flow-through pool anchor oligonucleotide (so as not to be confused with anchor/anchored reads in repeat amplification analysis). Under limiting dilution conditions, the adaptor-modified single-stranded template DNA is added to the flow-through pool and immobilized by hybridization with the anchor oligonucleotide. The ligated DNA is then... Fragment extension and bridging amplification are used to generate an ultra-high-density sequencing flow pool with hundreds of millions of clusters, each containing approximately 1,000 copies of the same template. In one implementation, randomly fragmented genomic DNA is amplified using PCR prior to cluster amplification. Alternatively, an amplification-free (e.g., PCR-free) genomic library is prepared, and cluster amplification is used only to enrich the randomly fragmented genomic DNA (Kozarewa et al., Nature Methods, Chapter 6: pp. 291–295 [2009]). The template is sequenced using a robust four-color DNA sequencing-by-synthesis technique employing a reversible terminator with removable fluorescent dyes. High-sensitivity fluorescence detection is achieved using laser excitation and total internal reflection optics. Short reads of approximately tens to hundreds of base pairs are aligned to a reference genome, and specially developed data analysis pipeline software is used to identify a unique mapping between the short reads and the reference genome. After the first read is completed, the template can be regenerated in situ to enable a second read from the opposite end of the fragment. Thus, sequencing can be performed using single-end or paired-end DNA fragments.

Various embodiments of the present invention may use sequencing-by-synthesis that allows for sequencing of paired ends. In some embodiments, Illumina's sequencing-by-synthesis includes clustering fragments. Clustering is a process in which each fragment molecule undergoes isothermal amplification. In some embodiments, as in the example described herein, the fragment has two different adaptors attached to both ends of the fragment, which allow the fragment to hybridize with two different oligonucleotides on the surface of the flow-through lane. The fragment also includes or is attached to two index sequences at both ends of the fragment, which provide markers to identify different samples in multiplex sequencing. In some sequencing platforms, the fragment to be sequenced is also referred to as an insert sequence.

In some implementations, the flow cell used for clustering on the Illumina platform is a glass slide with lanes. Each lane is a glass channel coated with primer motifs of two types of oligonucleotides. Hybridization is achieved by using the first of the two types of oligonucleotides on the surface. This oligonucleotide is complementary to a first linker at one end of the fragment. Polymerase produces the complementary strand of the hybridized fragment. The double-stranded molecule denatures, and the original template strand is washed away. The remaining strand, parallel to many other remaining strands, is used for clonal amplification via bridging applications.

In bridging amplification, one strand folds, and a second adaptor region at the second end of this strand hybridizes with a type II oligonucleotide on the surface of the flow cell. Polymerase produces a complementary strand, forming a double-stranded bridging molecule. This double-stranded molecule denatures, causing two single-stranded molecules to link to the flow cell via two different oligonucleotides. This process is then repeated repeatedly, simultaneously in millions of clusters, resulting in clonal amplification of all fragments. After bridging amplification, the reverse strand is cleaved and washed away, leaving only the forward strand. The 3' end is sealed to prevent unwanted initiation.

Following clustering, sequencing begins with extending the first sequencing primer to generate the first read. In each cycle, fluorescently labeled nucleotides are competitively added to the growing strand. Only one fluorescently labeled nucleotide is incorporated into the template-based sequence. After each nucleotide addition, the cluster is excited by a light source and emits a characteristic fluorescent signal. The number of cycles determines the read length. The emission wavelength and signal intensity determine the base determination. All identical strands of a given cluster are read simultaneously. Hundreds of millions of clusters are sequenced in massively parallel fashion. Upon completion of the first read, the read product is washed away.

In the next step of the protocol, which includes two index primers, index 1 primer is introduced and hybridized to the index 1 region on the template. The index region provides identification of the fragment, which can be used to demultiplex the sample during multiplexing. An index 1 read is generated, similar to the first read. After the index 1 read is complete, the read product is washed away, and the 3' end of the strand is deprotected. The template strand is then folded and bound to a second oligonucleotide on the flow cell. The index 2 sequence is read in the same manner as index 1. The index 2 read product is then washed away at the end of the step.

After reading two indices, read 2 is initiated by extending the second flow cell oligonucleotide using polymerase to form a double-stranded bridge. The double-stranded DNA is denatured, and the 3' end is blocked. The original forward strand is cleaved and washed away, leaving the reverse strand. Read 2 is then initiated with the introduction of the read 2 sequencing primer. As with read 1, the sequencing steps are repeated until the desired length is achieved. The read 2 product is washed away. This entire process generates millions of reads, representing all fragments. Sequences from the merged sample library are separated based on unique indices introduced during sample preparation. For each sample, reads with similar extensions determined by base determination are locally clustered. Forward and reverse reads are paired to generate adjacent sequences. These adjacent sequences are aligned to a reference genome for variant identification.

The above-described sequencing-by-synthesis examples include paired end reads, which are used in many embodiments of the methods disclosed in this invention. Paired end sequencing involves two reads from both ends of a fragment. When a pair of reads is mapped to a reference sequence, the base pair distance between the two reads can be determined, and this distance can then be used to determine the length of the fragment from which the reads are obtained. In some cases, a fragment spanning two groups will have one read in its paired end read aligned with one group and the other with an adjacent group. This will become less frequent as groups become longer or reads become shorter. Various methods can be used to indicate the group membership of these fragments. For example, these fragments can be omitted when determining the fragment size frequency of a group; these fragments can be counted for both in adjacent groups; these fragments can be assigned to a group containing a larger number of base pairs from both groups; or these fragments can be assigned to two groups with weights associated with a subset of base pairs in each group.

Paired-end reads can use insert sequences of different lengths (i.e., different fragment sizes to be sequenced). By default in this disclosure, paired-end reads are used to refer to reads obtained from various insert sequence lengths. In some cases, to distinguish short insert sequence paired-end reads from long insert sequence paired-end reads, the latter are also referred to as partner reads. In some embodiments including partner reads, two biotin-linked adaptors are first attached to the two ends of a relatively long insert sequence (e.g., several kb). The biotin-linked adaptors then connect the two ends of the insert sequence to form a circular molecule. Subfractions containing the biotin-linked adaptors can then be obtained by further fragmenting this circular molecule. The subfractions containing the two ends of the original fragment in reverse sequence order can then be sequenced using the same procedure as described above for short insert sequence paired-end sequencing. Further details regarding mated-end sequencing using the Illumina platform are available online at the following URL, which is incorporated herein by reference in its entirety: res|.|illumina|.|com/documents/products/technotes/technote_nextera_matepair_data_processing. Additional information on mated-end sequencing can be found in U.S. Patent 7,601,499 and U.S. Patent Publication 2012/0,053,063, the materials relating to mated-end sequencing methods and apparatus of which are incorporated by reference.

After DNA fragment sequencing, reads of a predetermined length (e.g., 100 bp) are mapped to or aligned with a known reference genome. The mapped or aligned reads and their corresponding positions on the reference sequence are also referred to as tags. In one implementation, the reference genome sequence is the NCBI36/hg18 sequence, available on the World Wide Web at genome dot ucsc dot edu/cgi-bin/hgGateway?org=Human&db=hg18&hgsid=166260105. Alternatively, the reference genome sequence is GRCh37/hg19, available on the World Wide Web at genome dot ucsc dot edu/cgi-bin/hgGateway. Other sources of public sequence information include GenBank, dbEST, dbSTS, EMBL (European Molecular Biology Laboratory), and DDBJ (Japan DNA Database). Various computer algorithms can be used for sequence alignment, including but not limited to BLAST (Altschul et al., 1990), BLITZ (MPsrch) (Sturrock and Collins, 1993), FASTA (Person and Lipman, 1988), BOWTIE (Langmead et al., Genome Biology, 10: R25.1-R25.10 [2009]), or ELAND (Illumina, San Diego, CA, USA). In one embodiment, one end of a clone amplified copy of a plasma cfDNA molecule is sequenced and processed by bioinformatics alignment analysis using an Illumina Genome Analyzer that utilizes the efficient, large-scale alignment of the Nucleotide Database (ELAND) software.

Other sequencing methods and systems can be used to obtain sequence reads.

Devices and systems for determining CNV

The analysis of sequencing data and the diagnosis derived therefrom are typically performed using various computer-executed algorithms and programs. Therefore, some embodiments employ processes that include data stored in or transmitted through one or more computer systems or other processing systems. The embodiments disclosed herein also include means for performing these operations. This means may be specifically configured for the desired purpose, or it may be a general-purpose computer (or group of computers) that can be selectively activated or reconfigured by computer programs and/or data structures stored in the computer. In some embodiments, the processor group performs some or all of the operations in the analysis operations collaboratively (e.g., via a network or cloud computing) and/or in parallel. The processors or processor groups used to perform the methods described herein can be of various types, including microcontrollers and microprocessors, such as programmable devices (e.g., CPLDs and FPGAs) and non-programmable devices (such as gate array ASICs or general-purpose microprocessors).

Furthermore, some embodiments involve tangible and/or non-transitory computer-readable media or computer program products, which include program instructions and/or data (including data structures) for performing various computer-implemented operations. Examples of computer-readable media include, but are not limited to, semiconductor memory devices, magnetic media (such as disk drives), magnetic tape, optical media (such as CDs), magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as read-only memory devices (ROM) and random access memory (RAM). The computer-readable medium may be directly controlled by an end user, or the medium may be indirectly controlled by an end user. Examples of directly controlled media include media located at the user's facility and/or media not shared with other entities. Examples of indirectly controlled media include media that the user can indirectly access via external networks and/or via services providing shared resources (such as the "cloud"). Examples of program instructions include machine code generated by a compiler and files containing higher-level code that can be executed by a computer using an interpreter.

In various embodiments, the data or information used in the methods and apparatus disclosed herein is provided in electronic format. Such data or information may include reads and tags derived from nucleic acid samples, counts or densities of such tags compared to specific regions of a reference sequence (e.g., compared to chromosomes or chromosomal segments), reference sequences (including those providing only or primarily polymorphism), chromosome and segment dosages, determinations (such as aneuploidy determinations), normalized chromosome and segment values, pairings of chromosomes or segments with corresponding normalized chromosomes or segments, consultations, diagnoses, etc. As used herein, data or other information provided in electronic format may be stored on a machine and transferred between machines. Conventionally, data in electronic format is provided digitally and may be stored as bits and/or bytes in various data structures, lists, databases, etc. Data may be represented electronically, optically, etc.

One embodiment provides a computer program product for generating output indicating the presence of aneuploidy (e.g., fetal aneuploidy) or cancer in a test sample. The computer product may contain instructions for performing any one or more of the methods described above for determining chromosomal abnormalities. As explained, the computer product may include a non-transitory and/or tangible computer-readable medium having computer-executable or compilable logic (e.g., instructions) recorded thereon for enabling a processor to determine chromosome dosage and, in some cases, the presence of fetal aneuploidy. In one example, the computer product includes a computer-readable medium having computer-executable or compilable logic (e.g., instructions) recorded thereon for enabling a processor to diagnose fetal aneuploidy, the computer-executable or compilable logic including: receiving sequencing data from at least a portion of nucleic acid molecules from a maternal biological sample, said sequencing data including calculated chromosome and/or segment dosages; computer-aided logic for analyzing fetal aneuploidy based on the received data; and an output program for generating output indicating the presence, absence, or type of said fetal aneuploidy.

Sequence information from the sample under consideration can be mapped to a chromosome reference sequence to identify the number of sequence tags for each chromosome in any one or more chromosomes of interest, and the number of sequence tags for the normalized segment sequences of each chromosome in any one or more chromosomes of interest. In various embodiments, for example, the reference sequence is stored in a database such as a relational database or an object database.

It should be understood that performing the computational operations of the methods disclosed herein is impractical, and in most cases impossible, for unassisted humans. For example, mapping a single 30bp read from a sample to any one chromosome in a human chromosome without the aid of a computational device could take years of effort. Of course, the problem is complex because reliable aneuploidy determination typically requires mapping thousands (e.g., at least about 10,000) or even millions of reads to one or more chromosomes.

The methods disclosed herein can be performed using a system for evaluating the copy number of a gene sequence of interest in a test sample. The system includes: (a) a sequencer for receiving nucleic acids from the test sample and providing nucleic acid sequence information from the sample; (b) a processor; and (c) one or more computer-readable media having instructions stored thereon for execution on the processor to perform a method for identifying any CNV (e.g., a chromosome or partial aneuploidy).

In some embodiments, the method is indicated by a computer-readable medium having computer-readable instructions stored thereon, the computer-readable instructions being used to perform a method for identifying any CNV (e.g., a chromosome or partial aneuploidy). Thus, one embodiment provides a computer program product comprising one or more computer-readable non-transitory storage media having computer-executable instructions stored thereon, the computer-executable instructions, when executed by one or more processors of a computer system, causing the computer system to implement a method for assessing the copy number of sequences of interest in a test sample containing cell-free fetal and maternal nucleic acids. The method includes: (a) receiving a sequence read obtained by sequencing a cell-free nucleic acid fragment in a test sample; (b) aligning the sequence read of the cell-free nucleic acid fragment to a reference genome containing the sequence of interest to provide a test sequence tag, wherein the reference genome is divided into multiple groups; (c) determining the size of the cell-free nucleic acid fragment present in the test sample; (d) weighting the test sequence tag based on the size of the cell-free nucleic acid fragment from which the tag is obtained; (e) calculating the coverage of the group based on the weighted tag in (d); and (f) identifying copy number variations in the sequence of interest based on the calculated coverage. In some embodiments, weighting the test sequence tag includes biasing the coverage towards the test sequence tag, which is obtained from a cell-free nucleic acid fragment having the size or size range characteristics of a genome in the test sample. In some embodiments, weighting the test sequence tag includes assigning a value of 1 to tags obtained from cell-free nucleic acid fragments of a size or size range, and assigning a value of 0 to other tags. In some implementations, the method further includes determining a value for a fragment size parameter within a grouping of a reference genome (including the sequence of interest). This fragment size parameter comprises a certain amount of free nucleic acid fragments in the test sample whose fragment size is shorter or longer than a threshold. Here, identifying copy number variations in the sequence of interest involves using the value of the fragment size parameter and the coverage calculated in (e). In some implementations, the system is configured to assess the copy number in the test sample using the various methods and procedures described above.

In some implementations, the instructions may also include automatically recording method-related information, such as chromosome dosage and the presence or absence of fetal chromosomal aneuploidy, in the patient's medical record of the human subject from whom the maternal test sample is provided. Patient records may be maintained by, for example, a laboratory, physician's office, hospital, health maintenance organization, insurance company, or personal medical record website. Furthermore, based on the results of processor-implemented analysis, the method may also include prescribing, initiating, and/or modifying treatment of the human subject from whom the maternal test sample is obtained. This may include performing one or more additional tests or analyses on additional samples obtained from the subject.

The methods disclosed in this invention can also be performed using a computer processing system adapted or configured to perform methods for identifying any CNV (e.g., chromosomes or partial aneuploidy). One embodiment provides a computer processing system adapted or configured to perform the methods described herein. In one embodiment, the apparatus includes a sequencing device adapted or configured to sequence at least a portion of nucleic acid molecules in a sample to obtain the type of sequence information described elsewhere herein. The apparatus may also include components for processing the sample. Such components are described elsewhere herein.

Sequences or data may be directly or indirectly input into a computer or stored on a computer-readable medium. In one embodiment, the computer system is directly coupled to a sequencing device that reads and/or analyzes nucleic acid sequences from a sample. Sequences or other information from such tools are provided via an interface in the computer system. Alternatively, sequences processed by the system are provided by a sequence storage source such as a database or other repository. Once available for processing, a storage device or mass storage device buffers or stores the nucleic acid sequences, at least temporarily. Furthermore, the storage device may store tag counts for various chromosomes or genomes, etc. The storage may also store various routines and/or programs for analyzing the presented sequence or mapping data. Such programs/routines may include programs for performing statistical analyses, etc.

In one example, a user places a sample in a sequencing device. Data is collected and/or analyzed via the sequencing device connected to a computer. Software on the computer allows for data collection and/or analysis. The data can be stored, displayed (via a monitor or other similar device), and/or transmitted to another location. The computer may be connected to the Internet, which is used to transfer data to a handheld device used by a remote user (e.g., a physician, scientist, or analyst). It should be understood that data may be stored and/or analyzed prior to transmission. In some embodiments, raw data is collected and transmitted to a remote user or device that will analyze and/or store the data. Transmission may be via the Internet, but may also be via satellite or other connections. Alternatively, data may be stored on a computer-readable medium, and this medium may be sent to the end user (e.g., via mail). Remote users may be located in the same or different geographic locations, including but not limited to buildings, cities, states, countries, or continents.

In some implementations, the method further includes acquiring data on multiple polynucleotide sequences (e.g., reads, tags, and/or reference chromosome sequences) and transmitting that data to a computer or other computing system. For example, the computer may be connected to laboratory equipment such as a sample acquisition device, nucleotide amplification device, nucleotide sequencing device, or hybridization device. The computer can then acquire applicable data collected by the laboratory equipment. Data may be stored on the computer at any step, e.g., during real-time acquisition, before transmission, during or concurrent with transmission, or after transmission. Data may be stored on a computer-readable medium retrievable from the computer. The acquired or stored data may be transmitted from the computer to a remote location, for example, via a local area network (LAN) or a wide area network (WAN) such as the Internet. At that remote location, various operations can be performed on the transmitted data, as described below.

In the systems, apparatuses, and methods disclosed herein, the types of electronically formatted data that can be stored, transmitted, analyzed, and/or manipulated are as follows:

Reads obtained by sequencing nucleic acids in test samples

Tags obtained by aligning reads with a reference genome or other reference sequences

Reference genome or sequence

Sequence tag density—the tag count or number of each region (usually a chromosome or chromosome segment) in two or more regions of a reference genome or other reference sequence.

The identity of a normalized chromosome or chromosome segment to a specific chromosome or chromosome segment of interest.

The chromosome or chromosome segment (or other region) obtained from the chromosome or segment of interest and the corresponding normalized chromosome or segment dose.

Thresholds for determining whether chromosome dose is affected, unaffected, or not determined.

Actual determination of chromosome dosage

Diagnosis (clinical condition associated with the diagnosis)

Recommendations for further testing derived from judgment and/or diagnosis

Treatment and/or monitoring plans derived from judgment and/or diagnosis

These various types of data can be acquired, stored, analyzed, and/or manipulated at one or more locations using different devices. The range of processing options is extensive. In the simplest case, all or most of this information is stored and used at the location where the test sample is processed (such as a physician's office or other clinical setting). In the most complex case, the sample is acquired at one location, processed and optionally sequenced at different locations, reads are compared and determined at one or more different locations, and a diagnosis, recommendation, and/or plan is prepared at yet another location (which could be the location where the sample was acquired).

In various implementations, a sequencing device generates reads, which are then transmitted to a remote site where they are processed to produce an aneuploidy determination. At this remote location, for example, the reads are aligned to a reference sequence to generate tags, which are counted and assigned to the chromosome or segment of interest. Also at this remote location, the counts are converted into doses using the associated normalized chromosome or segment. And still at this remote location, the doses are used to generate the aneuploidy determination.

During processing, the following operations can be performed at different locations:

Sample collection

Sample processing before sequencing

sequencing

Analyze sequence data and derive aneuploidy determination.

diagnosis

Report the diagnosis and/or judgment to the patient or healthcare provider.

Develop plans for further treatment, testing, and/or monitoring.

Implementation Plan

consult

Any one or more of these operations can be automated, as described elsewhere in this document. Typically, sequencing and analysis of sequence data, as well as the derivation of aneuploidy determination, will be performed computationally. Other operations can be performed manually or automatically.

Examples of locations where sample collection can be performed include healthcare worker offices, clinics, patients' homes (where sample collection tools or kits are provided), and mobile medical vehicles. Examples of locations where pre-sequencing sample processing can be performed include healthcare worker offices, clinics, patients' homes (where sample processing equipment or kits are provided), mobile medical vehicles, and facilities of aneuploidy analysis providers. Examples of locations where sequencing can be performed include healthcare worker offices, clinics, healthcare worker offices, clinics, patients' homes (where sample sequencing equipment and/or kits are provided), mobile medical vehicles, and facilities of aneuploidy analysis providers. Sequencing locations may have dedicated network connections for transmitting sequence data (typically reads) in electronic format. Such connections can be wired or wireless and have data and can be configured to send data to a site where it can be processed and/or aggregated before being transmitted to a processing site. Data aggregators may be maintained by health organizations such as health maintenance organizations (HMOs).

Analysis and/or derivation operations can be performed at any of the aforementioned locations, or alternatively at other remote sites dedicated to computation and/or service for analyzing nucleic acid sequence data. Such locations include, for example, clusters, such as general-purpose server clusters, facilities for aneuploid analysis services, etc. In some embodiments, the computing devices used to perform the analysis are leased or rented. Computing resources can be part of an internet-accessible set of processors, such as processing resources commonly referred to as the cloud. In some cases, computation is performed by parallel or massively parallel processor groups, either correlated or uncorrelated. Distributed processing such as cluster computing, grid computing, etc., can be used to accomplish this processing. In such embodiments, clusters or grids of computing resources collectively form a supercomputer consisting of multiple processors or computers that work together to perform the analysis and/or derivation described herein. These techniques, as well as more conventional supercomputers, can be used to process sequence data as described herein. Each technique is a form of parallel computing relying on processors or computers. In the case of grid computing, these processors (typically all computers) are connected via a network (private, public, or the internet) using conventional network protocols such as Ethernet. In contrast, supercomputers have many processors connected via a local high-speed computer bus.

In some implementations, a diagnosis (e.g., a fetus having Down syndrome or a patient having a specific type of cancer) is generated at the same location as the analytical procedure. In other implementations, the diagnosis is performed at a different location. In some examples, reporting a diagnosis is performed at the location where the sample was obtained, although this is not always the case. Examples of locations where diagnoses may be generated or reported and/or where program development may be performed include a healthcare professional's office, a clinic, a computer-accessible internet site, and handheld devices (such as mobile phones, tablets, smartphones, etc.) with wired or wireless connectivity to a network. Examples of locations where consultations may be performed include a healthcare professional's office, a clinic, a computer-accessible internet site, a handheld device, etc.

In some implementations, sample collection, sample processing, and sequencing are performed at a first location, and analysis and derivation are performed at a second location. However, in some cases, sample collection is performed at one location (e.g., a healthcare worker's office or clinic), and sample processing and sequencing are performed at different locations, optionally the same location where analysis and derivation are performed.

In several implementations, the sequence of operations listed above can be initiated by a user or entity to perform sample collection, sample processing, and/or sequencing. After one or more of these operations have been initiated, other operations can naturally follow. For example, a sequencing operation can cause reads to be automatically collected and sent to a processing device, which then typically and possibly automatically performs sequence analysis and aneuploidy deduction without further user intervention. In some implementations, the results of this processing operation (which may be reformatted for diagnosis) are then automatically delivered to system components or entities that process the information to healthcare professionals and/or patients. As explained, such information can also be automatically processed to generate treatment, testing, and/or monitoring plans, and possibly counseling information. Thus, initiating early operations can trigger an end-to-end sequence in which healthcare professionals, patients, or other stakeholders are provided with diagnostic, planning, counseling, and/or other information that can be used to take action based on the patient's condition. This can be achieved even if the parts of the system are physically separate and may be located, for example, away from the sample and sequencing devices.

Figure 17 illustrates a specific embodiment of a distributed system for generating a judgment or diagnosis from a test sample. Sample collection location 01 is used to obtain the test sample from a patient (such as a pregnant woman or a presumed cancer patient). The sample is then provided to processing and sequencing location 03, where the test sample can be processed and sequenced as described above. Location 03 includes means for processing the sample and means for sequencing the processed sample. As described elsewhere herein, the sequencing result is a collection of reads, which is typically provided in electronic format and made available to networks such as the Internet, indicated by reference numeral 05 in Figure 17.

The sequence data is provided to a remote location 07 for performing analysis and decision generation. This location may include one or more powerful computing devices, such as computers or processors. After the computing resources at location 07 have completed their analysis and generated a decision based on the received sequence information, the decision is relayed back to network 05. In some implementations, not only is a decision generated at location 07, but also an associated diagnosis. This decision and/or diagnosis is then transmitted over the network and returned to sample collection location 01, as shown in Figure 17. As explained, this is merely one of many variations regarding how to divide the various operations associated with generating decisions or diagnoses across various locations. A common variation involves providing sample collection, processing, and sequencing at a single location. Another variation involves providing processing and sequencing at the same location as analysis and decision generation.

Figure 18 illustrates the options for performing various operations at different locations. In the most detailed sense shown in Figure 18, each of the following operations is performed at a separate location: sample acquisition, sample processing, sequencing, read alignment, determination, diagnosis and reporting, and/or planning development.

In one implementation that aggregates some of these operations, sample processing and sequencing are performed in one location, while read alignment, determination, and diagnosis are performed in separate locations. See the portion identified by reference character A in Figure 18. In another embodiment, identified by character B in Figure 18, sample acquisition, sample processing, and sequencing are all performed at the same location. In this embodiment, read alignment and determination are performed at a second location. Finally, diagnosis and reporting, and/or program development, are performed at a third location. In an embodiment shown by character C in Figure 18, sample acquisition is performed at a first location, sample processing, sequencing, read alignment, determination, and diagnosis are performed together at a second location, and reporting and/or program development are performed at a third location. Finally, in an embodiment labeled D in Figure 18, sample acquisition is performed at a first location, sample processing, sequencing, read alignment, and determination are all performed at a second location, and diagnosis and reporting, and/or program management, are performed at a third location.

One embodiment provides a system for determining the presence of aneuploidy in a test sample containing fetal and maternal nucleic acids, the system comprising a sequencer for receiving a nucleic acid sample and providing fetal and maternal nucleic acid sequence information from the sample; and one or more processors configured to: (a) determine a fetal score of the test sample, wherein the fetal score of the test sample indicates the relative amount of fetal cell-free nucleic acid fragments in the test sample; (b) receive, by a computer system, sequence reads obtained by sequencing the cell-free nucleic acid fragments in the test sample; and (c) process, by the computer system, the sequence reads obtained by sequencing the cell-free nucleic acid fragments in the test sample. The sequence read of the cell-free nucleic acid fragment is aligned with a reference genome containing the sequence of interest to provide a sequence tag; (d) the coverage of the sequence tag of at least a portion of the reference genome is determined by a computer system; and (e) the test sample is determined to be within an exclusion region based on the coverage of the sequence tag determined in (d) and the fetal fraction determined in (a), wherein the exclusion region is defined by a limit of detection (LOD) curve for at least fetal fraction, wherein the fetal fraction LOD curve varies with coverage values and indicates the minimum fetal fraction required to achieve the detection criteria given different coverage values.

In some embodiments of any of the systems provided herein, the sequencer is configured to perform next-generation sequencing (NGS). In some embodiments, the sequencer is configured to perform massively parallel sequencing using sequencing-by-synthesis with reversible dye terminators. In other embodiments, the sequencer is configured to perform sequencing-by-ligation. In still other embodiments, the sequencer is configured to perform single-molecule sequencing.

In some implementations of any of the systems provided herein, one or more processors are programmed to perform the various methods described above.

Another aspect of this disclosure relates to a computer program product comprising a non-transitory machine-readable medium storing program code that, when executed by one or more processors of a computer system, causes the computer system to: (a) determine a fetal score of a test sample, wherein the fetal score of the test sample indicates the relative amount of cell-free fetal nucleic acid fragments in the test sample; (b) receive, by the computer system, a sequence read obtained by sequencing the cell-free nucleic acid fragments in the test sample; (c) align the sequence read of the cell-free nucleic acid fragments with a reference genome containing the sequence of interest, thereby providing a sequence tag; (d) determine, by the computer system, the coverage of the sequence tag for at least a portion of the reference genome; and (e) determine, based on the coverage of the sequence tag determined in (d) and the fetal score determined in (a), that the test sample is within an exclusion region, wherein the exclusion region is defined by at least a limit of detection (LOD) curve for fetal score, wherein the fetal score LOD curve varies with coverage values and indicates the minimum fetal score required to achieve the detection criteria given different coverage values.

In some implementations of the systems provided herein, the computer program product includes a non-transitory machine-readable medium storing program code that will be executed by one or more processors to perform the various methods described above.

experiment

Example 1

The purpose of this embodiment is to illustrate how to obtain LOD curves using the above method.

Research Design

For this study, a portion of the data generated in the VeriSeq NIPT Accuracy Study (BRIGID-0147 VeriSeq NIPT: Accuracy Study Protocol and BRIGID-0166 VeriSeq NIPT—Accuracy Study Report) will be used to determine the detection limit of the VeriSeq NIPT solution system.

In the VeriSeq NIPT precision study, pooled cfDNA from 21-triploid and non-pregnant plasma pools was combined to generate a 5% fetal fraction 21-triploid affected pool, which was processed by VeriSeq NIPT assay. Male maternal pools were also processed unchanged by VeriSeq NIPT assay.

Table 1 shows the in-laboratory precision component of the precision study design used for LOD studies. The in-laboratory precision component consisted of two Hamilton instruments, two NextSeq instruments, and three reagent batches, running a total of 12 times over 6 days.

Table 1—Laboratory Accuracy Component of Accuracy Studies

Pair automated and sequencing instruments with consistent operators. Combine operator and instrument variants.

Data generated using reagent batches 1 and 2 were used to establish the limits of detection. Synthetic sequence data from T21 samples (5% FF) were generated by combining sequence data from non-pregnant women with different fetal fractions. This process was repeated for five different sequencing depths (2 million, 4 million, 6 million, 8 million, and 10 million unique aligned read coverage—see Table 2). Dilution points were selected on a wide range of fine grids to cover the expected limits of detection for all sequencing depths (1.25%–4.5%), with a step size of 0.25%.

For each dilution point and each coverage level, 20 replicate synthetic samples were generated (see DEV report-0072 for a more detailed method description, Mathematical Model for Log Likelihood Ratio Score and Detection Limit Prediction).

Using computer-generated dilution series, the LoD of T21 was established for each coverage level using a probabilistic unit regression method as described in BRIGID-0150 (VeriSeq NIPT Solution Design Validation Protocol Detection Limit). Furthermore, the LODs of T13 and T18 were determined using the LOD results of T21 and the known relationships between LOD T21 and LODs T13 and T18, as described in DEV Report-0072.

result

The quantitative score, log-likelihood ratio (LLR), and estimate of the fetal score for each repeated run were calculated using R&D Clinical Aneuploidy Detection and Analysis Suite (cADAS) v3.2.

Sample filtration and outliers

All template-free control (NTC) samples were excluded from the analysis. These samples were then removed, resulting in 564 samples, with N=48 in pool T21 and N-48 in the non-pregnant pool. These samples were further used for computer-simulated dilution analysis. The remaining samples were not used directly but were used as packing material for the analysis. One non-pregnant sample was marked as NES_FF_QC failure; this QC failure was ignored, and the sample was included in the analysis because it was non-pregnant (containing no fetal DNA) and therefore expected to fail in that metric.

Computer simulation of dilution quality verification

The affected sample was prepared from a combined mixture of male and female plasma samples, with each sample comprising approximately 50%. Therefore, Y chromosome DNA was present in the sample and could be used to validate the computer-simulated dilution process. Figure 19 shows the Y chromosome coverage (left) and FF fraction estimates (right) of the synthesized sample based on the dilution fraction. Both linear plots show the reconciliation values between the FF (initial non-pregnant FF = 0%, and affected FF = 5%) and the computer-simulated dilution.

Determine the detection limit of the T21 triploid test

The procedure for determining the LOD based on experimental observations of a specified coverage level is described in BRIGID 0150vA (VeriSeq NIPT solution design validation protocol detection limit).

The measured LODs for T21 for each batch and each coverage level are summarized in Table 2.

Table 2 Detection Limits

Detection limit contrast coverage

The theoretical model described in DEV Report-0072 (the mathematical model for predicting the log-likelihood ratio score and detection limit) predicts a linear LOD contrast coverage on a Log-Log scale. Figure 20 shows the results of the linear fit to the LOD contrast coverage. We observed a very good match between the experimental data and the predicted behavior.

The fitted LOD, compared to the coverage line, predicts the detection limit for any coverage. The result of overlaying the predicted LOD with the observed LOD is shown in Figure 21.

Detection limit in average case

The total population of samples has variable coverage. Figure 22 shows the coverage distribution (NES) of a large number of samples (N = 14,400) processed using the VeriSeq NIPT assay, version V2. The expected LOD of the total population was calculated based on the expected LOD of coverage within the coverage distribution range.

Where LOD(NES) is the LOD based on coverage, and p(NES) is the probability density function of NES shown in Figure 22. The average LOD obtained for T21 is shown in Table 3.

Table 3 Detection Limits for Average Population Coverage

measure Batch 1 Batch 2 Batch 3 LOD21 2.81% 2.40% 2.43%

Detection limits for other aneuploid T13 and T18

As described in previous research (Mathematical Model for Log-Likelihood Ratio Score and Detection Limit Prediction), the detection limit for aneuploidy on different chromosomes can be inferred from the known LOD of at least one chromosome. Using relations from previous research, we were able to infer the LOD of T13 and T18 triploids using previously determined doubling factors, and the results are shown in Tables 4 and 5.

Table 4. Detection limits for the average population coverage inferred from the acceptance criteria in T13 and T18 assessments.

aneuploidy Observe LoD LoD coefficient LoD inferred from T21 T21 2.81% 100% 2.81% T18 NA 81% 2.27% T13 NA 72% 2.02%

Table 5 Accuracy %CV Acceptance Criteria

measure %FF Through standards Pass/Fail LOD13 2.02% LOD < 4% pass LOD18 2.27% LOD < 4% pass LOD21 2.81% LOD < 4% pass

in conclusion

The limits of detection (LODs) for the VeriSeq NIPT system for fetal fractions of fetal cfDNA (triploid 13/18/21) in a normal maternal diploid cfDNA background have been determined. The LODs are summarized in Table 6. For all aneuploidy tested, the observed LODs were found to be lower than the values specified in the testing requirements document. Based on the data presented in Table 6, the determined LODs met the acceptance criteria of another study targeting all aneuploidy.

Table 6: Fractional LOD of aneuploid fetuses

aneuploidy LOD(%FF) T13 2.02% T18 2.27% T21 2.81%

Example 2

Example 2 illustrates empirical data on various performance metrics of the conventional method labeled as the NES and fetal fractional LOD curve method described above.

Figure 23 shows NES coverage based on observed fetal scores. A two-step coverage thresholding technique was used to exclude samples. Various samples falling into the exclusion zone are shown in the figure. Most excluded samples had fetal scores between 0% and 20%, and their coverage was limited by two threshold levels.

Figure 24 illustrates data exclusion using exclusion regions defined by LOD curves and read thresholds. The left figure shows a read threshold of 2 million reads. The right figure shows data using a coverage threshold of 1 million reads. As shown, the coverage threshold can be lowered to exclude fewer samples when fewer samples are desired. Many excluded samples have relatively high read coverage and relatively low observed fetal scores. In contrast, in Figure 23, many excluded samples have relatively high fetal scores and relatively low coverage.

Figure 25 shows the pass and failure rates for the first and second runs of the existing method and the LOD QC method described above. The data shows that both methods have similar pass and failure rates. The final failure rate of the existing method is 0.42%, while the final failure rate of the LOD QC method is 0.38%.

Figure 26 shows data excluded by a two-step thresholding method and salvaged by the fetal fractional depth-of-reading (LOD) curve method, labeled 2604. Samples excluded by the LOD QC method and salvaged by conventional methods are shown in region 2602. The left panel of Figure 26 shows data with a coverage threshold of 2 million reads combined with the LOD curve. The right panel shows salvaged data with a lower coverage threshold of 1 million reads. Using the 2 million read coverage threshold in the left panel, the LOD QC method excluded 61 additional samples. However, when the coverage threshold was lowered to 1 million reads, 151 fewer samples were excluded by the LOD QC method.

Figure 27 shows samples saved using the LOD QC method. The left figure shows samples excluded in region 2702 by conventional methods and saved by the LOD QC method. After rerun, 80% of these samples fell into the containment region 2704. 10% remained in the same region 2702. 9.7% of the samples fell below the LOD curve and entered the exclusion region. The right figure shows samples saved by applying a 1 million read threshold to the LOD curve. Region 2708 shows samples saved by a 1 million read threshold after being excluded by conventional methods. After rerunning the samples, 80% moved into the containment region 2710, 13.7% remained in the same region 2708, and 6.3% fell below the LOD curve and entered the exclusion region. This figure demonstrates that the LOD method alone, or the LOD method in combination with a coverage threshold method, can save a large percentage of samples that would otherwise be excluded. During implementation, the LOD QC method helps avoid the need to rerun the salvaged sample, thus saving time, cost, and resources for CNV detection.

Figure 28 shows the data excluded by the LOD QC method in region 2802. After rerun, a few samples moved above the LOD curve into region 2708. More data remained in the same region 2802 or moved further below the LOD curve 2708. The results from Figures 27 and 28 indicate that the LOD QC method includes many samples that actually meet the QC criteria and excludes samples that do not meet the QC criteria, even during rerun. This demonstrates that the LOD QC method can intelligently and correctly separate positive and negative samples.

Figure 29 shows the pass rate and failure rate of the existing method and the LOD QC method in two runs. Overall, the pass rate and failure rate are similar for both methods in the first and second runs.

Figure 30 shows that the 75% confidence level LOD curve rescued sample 3002, which was a false positive for 21-triploidy (T21). The curve also rescued sample 3004, which was a false negative for 18-triploidy (T18). This resulted in a 100% reduction in T21 false positives and a 50% reduction in T18 false negatives. This demonstrates that the LOD QC method can increase both sensitivity and specificity.

Figure 31 shows that, for the simulated T21 sample, samples passing the LOD QC can be detected with a sensitivity of 99.7%, while samples passing the LOD QC can be detected with a sensitivity of 88.1%. Regardless of the LOD QC status, the overall sensitivity for all samples is 99.7.

Claims (75)

1. A method for processing test samples containing cell-free nucleic acid fragments derived from the mother and fetus, the method being implemented using a computer system including a memory and one or more processors, the method comprising: (a) Determine an observation of the fetal fraction of the test sample, wherein the fetal fraction of the test sample indicates the relative amount of fetal cell-free nucleic acid fragments in the test sample; (b) The computer system receives a sequence read obtained by sequencing the free nucleic acid fragment in the test sample; (c) The computer system aligns the sequence read of the free nucleic acid fragment with a reference genome containing the sequence of interest, thereby providing a sequence tag; (d) The computer system determines the coverage of the sequence tag for at least a portion of the reference genome; (e) Determining the test sample within the exclusion region based on the sequence tag coverage determined in (d) and the observed fetal fraction determined in (a), wherein the exclusion region is defined by at least a fetal fraction detection limit (LOD) curve, wherein the fetal fraction LOD curve varies with coverage values and indicates the minimum observed fetal fraction required to achieve a given confidence level for a given coverage value, the given confidence level being the minimum fetal fraction greater than the corresponding true fetal fraction required to achieve the detection standard; and (f) As a response to determining that the test sample is within the exclusion region, the test sample is excluded for determining the copy number variation (CNV) of the sequence of interest, or the test sample is re-sequencing to obtain a re-sequencing read for determining the CNV of the sequence of interest. The LOD curve mentioned above is obtained through the following: (i) For each of a plurality of observed fetal scores and for each of a plurality of coverage levels, select one or more selected training samples from a first plurality of training samples, wherein the true fetal scores of the training samples have a given confidence level, wherein each training sample includes cell-free nucleic acid fragments derived from the mother and the fetus. (ii) Using one or more selected training samples, obtain curves of real fetal scores versus observed fetal scores at multiple coverage levels; (iii) Using a second plurality of training samples to obtain real fetal scores, the real fetal scores having at least a minimum fetal score required to achieve the detection criteria for a plurality of coverage levels; (iv) Using the true fetal score versus observed fetal score curve and the true fetal score having at least the minimum fetal score required to achieve the detection criteria, obtain multiple minimum observed fetal scores for multiple coverage levels required with a given confidence level, said given confidence level being the minimum true fetal score having at least the minimum fetal score required to achieve the detection criteria; and (v) Obtain fetal score LOD curves using multiple minimum observed fetal scores for multiple coverage levels. 2. The method according to claim 1, further comprising: Use the resequencing sequence reads to repeat (a)-(d); and The test sample was determined to be outside the exclusion zone. 3. The method of claim 1, wherein the fetal fractional LOD curve is obtained based on the LOD of the affected training samples affected by CNV. 4. The method of claim 3, wherein the affected training sample includes a computer-simulated sample. 5. The method of claim 3, wherein the affected training sample comprises an in vitro sample. 6. The method of claim 3, wherein the affected training sample is obtained by combining samples having two or more fetal scores. 7. The method of claim 1, wherein the detection criterion is that, at the minimum observed value of the fetal score and the given coverage value, there exists a given confidence level, wherein the given confidence level is greater than a specified LOD required to detect the affected sample at a certain detection confidence level. 8. The method of claim 7, wherein the detection standard is an X% confidence level, wherein the X% confidence level is, with respect to the minimum observed value of the fetal score, the corresponding true fetal score is greater than the specified LOD required to detect the affected sample at a Y% detection confidence level. 9. The method of claim 8, wherein X% is about 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, 99%, or 99.5%. 10. The method of claim 8, wherein Y% is a detection confidence level of about 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, or 99%. 11. The method of claim 8, wherein X% is 50% and Y% is 95%. 12. The method of claim 7, wherein the designated LOD is determined as the minimum observed fetal fraction of the affected sample in which Y% can be detected. 13. The method of claim 7, wherein the distribution of the true fetal scores observed at the observation coverage is used to obtain the detection criteria for the observed fetal scores at said coverage. 14. The method according to any one of claims 1-13, wherein the exclusion region is below the fetal fractional level of deviation (LOD) curve. 15. The method according to any one of claims 1-13, wherein the exclusion region is defined by the fetal fractional LOD curve and the coverage threshold. 16. The method according to any one of claims 1-13, wherein the exclusion region is below both the fetal fraction LOD curve and the coverage threshold. 17. The method according to any one of claims 1-13, wherein determining the sequence tag coverage of the at least portion of the reference genome comprises: (i) Divide the reference genome into multiple groups; (ii) Determine the number of sequence tags to be compared with each group; and (iii) The sequence tag coverage is determined using the number of sequence tags in the groupings of at least a portion of the reference genome. 18. The method of claim 17, further comprising, prior to (iii), adjusting the number of sequence tags compared with the group by taking into account inter-group variation due to factors other than copy number variation. 19. The method according to any one of claims 1-13, wherein the observed value of the fetal fraction of the test sample is determined based on the size of the free nucleic acid fragment. 20. The method of claim 19, wherein the observed value of the fetal fraction of the test sample is determined by the following steps: Obtain the frequency distribution of the size of the free nucleic acid fragments; and The frequency distribution is applied to a model that correlates fetal fractions with the frequency of fragment size to obtain observations of the fetal fractions. 21. The method of claim 17, wherein the observed value of the fetal score of the test sample is determined based on the coverage information of the grouping of the reference genome. 22. The method of claim 21, wherein the observed fetal score is calculated by applying coverage values of a plurality of groups of the reference genome to a model that correlates the fetal score with the coverage of the groups to obtain the observed fetal score. 23. The method of claim 22, wherein the plurality of groups of the reference genome have a higher fraction of fetal cell-free nucleic acid fragments than the other groups. 24. The method according to any one of claims 1-13, wherein the observed value of the fetal score of the test sample is determined based on the coverage information of the sex chromosome grouping. 25. A system for assessing the copy number of a nucleic acid sequence of interest in a test sample, the system comprising a processor and one or more computer-readable storage media having instructions stored thereon for performing the following operations on the processor: (a) Determine an observation of the fetal fraction of the test sample, wherein the fetal fraction of the test sample indicates the relative amount of fetal cell-free nucleic acid fragments in the test sample; (b) Receiving a sequence read obtained by sequencing the free nucleic acid fragment in the test sample; (c) Aligning the sequence read of the free nucleic acid fragment with a reference genome containing the sequence of interest to provide a sequence tag; (d) Determine the coverage of the sequence tag for at least a portion of the reference genome; and (e) The test sample is determined to be within an exclusion region based on the sequence tag coverage determined in (d) and the observed fetal fraction determined in (a), wherein the exclusion region is defined by at least a fetal fraction detection limit (LOD) curve, wherein the fetal fraction LOD curve varies with coverage values and indicates the minimum observed fetal fraction required to achieve a given confidence level for a given coverage value, wherein the given confidence level is the minimum fetal fraction greater than the corresponding true fetal fraction required to achieve the detection criteria. The LOD curve mentioned above is obtained through the following: (i) From a first plurality of training samples, for each of a plurality of observed fetal scores and for each of a plurality of coverage levels, one or more selected training samples are selected, the true fetal scores of the training samples having a given confidence level, wherein each training sample includes cell-free nucleic acid fragments derived from the mother and fetus. (ii) Using one or more selected training samples, obtain curves of real fetal scores versus observed fetal scores at multiple coverage levels; (iii) Using a second plurality of training samples to obtain real fetal scores, the real fetal scores having at least a minimum fetal score required to achieve the detection criteria for a plurality of coverage levels; (iv) Using the true fetal score versus observed fetal score curve and the true fetal score having at least the minimum fetal score required to achieve the detection criteria, obtain multiple minimum observed fetal scores for multiple coverage levels required with a given confidence level, said given confidence level being the minimum true fetal score having at least the minimum fetal score required to achieve the detection criteria; and (v) Obtain fetal score LOD curves using multiple minimum observed fetal scores for multiple coverage levels. 26. A machine-readable medium storing program code, which, when executed by one or more processors of a computer system, causes the computer system to: (a) Determine the observation of the fetal fraction of the test sample, wherein the fetal fraction of the test sample indicates the relative amount of fetal cell-free nucleic acid fragments in the test sample; (b) Receiving a sequence read obtained by sequencing the free nucleic acid fragment in the test sample; (c) Aligning the sequence read of the free nucleic acid fragment with a reference genome containing the sequence of interest to provide a sequence tag; (d) Determine the coverage of the sequence tag for at least a portion of the reference genome; and (e) The test sample is determined to be within an exclusion region based on the sequence tag coverage determined in (d) and the observed fetal fraction determined in (a), wherein the exclusion region is defined by at least a fetal fraction detection limit (LOD) curve, wherein the fetal fraction LOD curve varies with coverage values and indicates the minimum observed fetal fraction required to achieve a given confidence level for a given coverage value, wherein the given confidence level is the minimum fetal fraction greater than the corresponding true fetal fraction required to achieve the detection criteria. The LOD curve mentioned above is obtained through the following: (i) From a first plurality of training samples, for each of a plurality of observed fetal scores and for each of a plurality of coverage levels, one or more selected training samples are selected, the true fetal scores of the training samples having a given confidence level, wherein each training sample includes cell-free nucleic acid fragments derived from the mother and fetus. (ii) Using one or more selected training samples, obtain curves of real fetal scores versus observed fetal scores at multiple coverage levels; (iii) Using a second plurality of training samples to obtain real fetal scores, the real fetal scores having at least a minimum fetal score required to achieve the detection criteria for a plurality of coverage levels; (iv) Using the true fetal score versus observed fetal score curve and the true fetal score having at least the minimum fetal score required to achieve the detection criteria, obtain multiple minimum observed fetal scores for multiple coverage levels required with a given confidence level, said given confidence level being the minimum true fetal score having at least the minimum fetal score required to achieve the detection criteria; and (v) Obtain fetal score LOD curves using multiple minimum observed fetal scores for multiple coverage levels. 27. A system for assessing the copy number of a nucleic acid sequence of interest in a test sample, the system comprising one or more processors and one or more computer-readable storage media having program code stored thereon for performing the following: (a) Determine an observation of the fetal fraction of the test sample, wherein the fetal fraction of the test sample indicates the relative amount of fetal cell-free nucleic acid fragments in the test sample; (b) Receiving a sequence read obtained by sequencing the free nucleic acid fragment in the test sample; (c) Aligning the sequence read of the free nucleic acid fragment with a reference genome containing the sequence of interest to provide a sequence tag; (d) Determine the coverage of the sequence tag for at least a portion of the reference genome; (e) Determining the test sample within the exclusion region based on the sequence tag coverage determined in (d) and the observed fetal fraction determined in (a), wherein the exclusion region is defined by at least a fetal fraction detection limit (LOD) curve, wherein the fetal fraction LOD curve varies with coverage values and indicates the minimum observed fetal fraction required to achieve a given confidence level for a given coverage value, the given confidence level being the minimum fetal fraction greater than the corresponding true fetal fraction required to achieve the detection standard; and (f) As a response to determining that the test sample is within the exclusion region, the test sample is excluded for determining the copy number variation (CNV) of the sequence of interest, or a re-sequencing read from the re-sequencing of the test sample is obtained for determining the CNV of the sequence of interest. The LOD curve mentioned above is obtained through the following: (i) From a first plurality of training samples, for each of a plurality of observed fetal scores and for each of a plurality of coverage levels, one or more selected training samples are selected, the true fetal scores of the training samples having a given confidence level, wherein each training sample includes cell-free nucleic acid fragments derived from the mother and fetus. (ii) Using one or more selected training samples, obtain curves of real fetal scores versus observed fetal scores at multiple coverage levels; (iii) Using a second plurality of training samples to obtain real fetal scores, the real fetal scores having at least a minimum fetal score required to achieve the detection criteria for a plurality of coverage levels; (iv) Using the true fetal score versus observed fetal score curve and the true fetal score having at least the minimum fetal score required to achieve the detection criteria, obtain multiple minimum observed fetal scores for multiple coverage levels required with a given confidence level, said given confidence level being the minimum true fetal score having at least the minimum fetal score required to achieve the detection criteria; and (v) Obtain fetal score LOD curves using multiple minimum observed fetal scores for multiple coverage levels. 28. The system of claim 27, wherein the program code includes code for implementing the following: Use the resequencing reads to repeat (a)-(d); Determine that the test sample is outside the exclusion zone; and A CNV determination is made for the test sample regarding the sequence of interest. 29. The system of claim 27, wherein the fetal fractional LOD curve is obtained based on the LOD of the affected training samples affected by CNV. 30. The system of claim 29, wherein the affected training sample includes a computer-simulated sample. 31. The system of claim 29, wherein the affected training sample comprises an in vitro sample. 32. The system of claim 29, wherein the affected training sample is obtained by combining samples having two or more fetal scores. 33. The system of claim 27, wherein the detection criterion is that, at the minimum observed value of the fetal score and the given coverage value, there exists a given confidence level, wherein the given confidence level is greater than a specified LOD required to detect the affected sample at a certain detection confidence level. 34. The system of claim 33, wherein the detection standard is an X% confidence level, wherein the X% confidence level is, with respect to the minimum observed value of the fetal score, the corresponding true fetal score being greater than the specified LOD required to detect the affected sample at a Y% detection confidence level. 35. The system of claim 34, wherein X% is about 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, 99%, or 99.5%. 36. The system of claim 34, wherein Y% is a detection confidence level of about 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, or 99%. 37. The system of claim 34, wherein X% is 50% and Y% is 95%. 38. The system of claim 33, wherein the designated LOD is determined as the minimum observed fetal fraction in which Y% of the affected sample can be detected. 39. The system of claim 33, wherein the distribution of the true fetal scores observed at the observation coverage is used to obtain the detection criteria for the observed fetal scores at the observation coverage. 40. The system according to any one of claims 27-39, wherein the exclusion region is below the fetal fractional OOD curve. 41. The system according to any one of claims 27-39, wherein the exclusion region is defined by the fetal fractional LOD curve and the coverage threshold. 42. The system according to any one of claims 27-39, wherein the exclusion region is below both the fetal fraction LOD curve and the coverage threshold. 43. The system according to any one of claims 27-39, wherein determining the sequence tag coverage of the at least portion of the reference genome comprises: (i) Divide the reference genome into multiple groups; (ii) Determine the number of sequence tags to be compared with each group; and (iii) The sequence tag coverage is determined using the number of sequence tags in the groupings of at least a portion of the reference genome. 44. The system of claim 43, wherein determining the sequence tag coverage of the portion of the reference genome further includes, prior to (iii), adjusting the number of sequence tags aligned with the group by taking into account inter-group variations due to factors other than copy number variations. 45. The system according to any one of claims 27-39, wherein the observed value of the fetal fraction of the test sample is determined based on the size of the free nucleic acid fragment. 46. The system of claim 45, wherein the observed value of the fetal fraction of the test sample is determined by the following steps: Obtain the frequency distribution of the size of the free nucleic acid fragments; and The frequency distribution is applied to a model that correlates fetal fractions with the frequency of fragment size to obtain observations of the fetal fractions. 47. The system of claim 43, wherein the observed value of the fetal score of the test sample is determined based on the coverage information of the grouping of the reference genome. 48. The system of claim 47, wherein the observed fetal score is calculated by applying coverage values of a plurality of groups of the reference genome to a model that correlates the fetal score with the coverage of the groups to obtain the observed fetal score. 49. The system of claim 48, wherein the plurality of groups of the reference genome have a higher fraction of fetal cell-free nucleic acid fragments than the other groups. 50. The system according to any one of claims 27-39, wherein the observed value of the fetal score of the test sample is determined based on coverage information of sex chromosome grouping. 51. A machine-readable medium storing program code, which, when executed by one or more processors of a computer system, causes the computer system to perform a method comprising the steps of: (a) Determine the observation of the fetal fraction of the test sample, wherein the fetal fraction of the test sample indicates the relative amount of fetal cell-free nucleic acid fragments in the test sample; (b) The computer system receives a sequence read obtained by sequencing the free nucleic acid fragment in the test sample; (c) The computer system aligns the sequence read of the cell-free nucleic acid fragment with a reference genome containing the sequence of interest, thereby providing a sequence tag; (d) The computer system determines the coverage of the sequence tag for at least a portion of the reference genome; (e) Determining the test sample within the exclusion region based on the sequence tag coverage determined in (d) and the observed fetal fraction determined in (a), wherein the exclusion region is defined by at least a fetal fraction detection limit (LOD) curve, wherein the fetal fraction LOD curve varies with coverage values and indicates the minimum observed fetal fraction required to achieve a given confidence level for a given coverage value, the given confidence level being the minimum fetal fraction greater than the corresponding true fetal fraction required to achieve the detection standard; and (f) As a response to determining that the test sample is within the exclusion region, the test sample is excluded for determining the copy number variation (CNV) of the sequence of interest, or a re-sequencing read from the re-sequencing of the test sample is obtained for determining the CNV of the sequence of interest. The LOD curve mentioned above is obtained through the following: (i) From a first plurality of training samples, for each of a plurality of observed fetal scores and for each of a plurality of coverage levels, one or more selected training samples are selected, the true fetal scores of the training samples having a given confidence level, wherein each training sample includes cell-free nucleic acid fragments derived from the mother and fetus. (ii) Using one or more selected training samples, obtain curves of real fetal scores versus observed fetal scores at multiple coverage levels; (iii) Using a second plurality of training samples to obtain real fetal scores, the real fetal scores having at least a minimum fetal score required to achieve the detection criteria for a plurality of coverage levels; (iv) Using the true fetal score versus observed fetal score curve and the true fetal score having at least the minimum fetal score required to achieve the detection criteria, obtain multiple minimum observed fetal scores for multiple coverage levels required with a given confidence level, said given confidence level being the minimum true fetal score having at least the minimum fetal score required to achieve the detection criteria; and (v) Obtain fetal score LOD curves using multiple minimum observed fetal scores for multiple coverage levels. 52. The machine-readable medium of claim 51, wherein the test sample has been determined to be negative for the CNV determination of the sequence of interest. 53. The machine-readable medium of claim 51, further comprising the step of: Use the resequencing reads to repeat (a)-(d); Determine that the test sample is outside the exclusion zone; and A CNV determination is made for the test sample regarding the sequence of interest. 54. The machine-readable medium of claim 51, wherein the fetal fractional LOD curve is obtained based on the LOD of the affected training sample influenced by CNV. 55. The machine-readable medium of claim 54, wherein the affected training sample comprises a computer-simulated sample. 56. The machine-readable medium of claim 54, wherein the affected training sample comprises an in vitro sample. 57. The machine-readable medium of claim 54, wherein the affected training sample is obtained by combining samples having two or more fetal scores. 58. The machine-readable medium of claim 51, wherein the detection criterion is that, at the minimum observed value of the fetal score and the given coverage value, there exists a given confidence level, wherein the given confidence level is greater than a specified LOD required to detect the affected sample at a certain detection confidence level. 59. The machine-readable medium of claim 58, wherein the detection standard is an X% confidence level, the X% confidence level being, with respect to the minimum observed value of the fetal score, a corresponding true fetal score greater than a specified LOD required to detect the affected sample at a Y% detection confidence level. 60. The machine-readable medium of claim 59, wherein X% is about 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, 99%, or 99.5%. 61. The machine-readable medium of claim 59, wherein Y% is a detection confidence level of about 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, or 99%. 62. The machine-readable medium of claim 59, wherein X% is 50% and Y% is 95%. 63. The machine-readable medium of claim 58, wherein the designated LOD is determined as the minimum observed fetal fraction in which Y% of the affected sample can be detected. 64. The machine-readable medium of claim 58, wherein the distribution of the actual fetal scores observed at the observation coverage is used to obtain the detection criteria for the observed fetal scores at said coverage. 65. The machine-readable medium according to any one of claims 51-64, wherein the exclusion region is below the fetal fractional level of distance (LOD) curve. 66. The machine-readable medium according to any one of claims 51-64, wherein the exclusion region is defined by the fetal fractional LOD curve and the coverage threshold. 67. The machine-readable medium according to any one of claims 51-64, wherein the exclusion region is below both the fetal fractional level of diameter (LOD) curve and the coverage threshold. 68. The machine-readable medium according to any one of claims 51-64, wherein the sequence tag coverage for determining the at least portion of the reference genome comprises: (i) Divide the reference genome into multiple groups; (ii) Determine the number of sequence tags to be compared with each group; and (iii) The sequence tag coverage is determined using the number of sequence tags in the groupings of at least a portion of the reference genome. 69. The machine-readable medium of claim 68, wherein determining the sequence tag coverage of the portion of the reference genome further includes, prior to (iii), adjusting the number of sequence tags aligned with the group by taking into account inter-group variations due to factors other than copy number variations. 70. The machine-readable medium according to any one of claims 51-64, wherein the observed value of the fetal fraction of the test sample is determined based on the size of the free nucleic acid fragment. 71. The machine-readable medium of claim 70, wherein the observed fetal fraction of the test sample is determined by the following steps: Obtain the frequency distribution of the size of the free nucleic acid fragments; and The frequency distribution is applied to a model that correlates fetal fractions with the frequency of fragment size to obtain observations of the fetal fractions. 72. The machine-readable medium of claim 68, wherein the observed value of the fetal score of the test sample is determined based on the coverage information of the grouping of the reference genome. 73. The machine-readable medium of claim 72, wherein the observed fetal score is calculated by applying coverage values of a plurality of groups of the reference genome to a model that correlates the fetal score with the coverage of the groups to obtain the observed fetal score. 74. The machine-readable medium of claim 73, wherein the plurality of groups of the reference genome have a higher fraction of fetal cell-free nucleic acid fragments than the other groups. 75. The machine-readable medium according to any one of claims 51-64, wherein the observed value of the fetal score of the test sample is determined based on coverage information of sex chromosome grouping.