CN105986007B - Detection method of cancer tumor suppressor gene cluster (TSG) - Google Patents

Abstract

In order to quickly and accurately obtain tumor suppressor genes that play a common role in the occurrence of cancer, the inventors designed a new method: first, by simulating random sampling, the pairwise relationship between all genes in the same region is searched for ; secondly, select genes that have a co-occurrence relationship; then, link these two-co-occurring genes to form a chain, and the genes in the chain must have a co-occurrence relationship; finally, combine the clinical information and expression of the sample data to verify whether genes in a chain play a common role in cancer development, and whether this common role is stronger than that of individual genes.

Description

A cancer tumor suppressor gene cluster (TSG) detection method

technical field

The present invention relates to the field of biological information. More specifically, the present invention relates to the detection of cancer tumor suppressor gene clusters.

Background technique

The genetic mutation theory of cancer has always occupied the mainstream of cancer mechanism theory. Since the identification of the first human gene HRAS mutation that may cause cancer in 1982, cancer research has entered the era of exploration and identification of oncogenes and tumor suppressor genes. Early studies focused on a single gene mutation in a genomic region that acts as a "driver" in tumorigenesis, but recent studies have found a large number of tumor suppressor genes with frequent deletion mutations in the genome, which are associated with heterozygous Deletion-related deletion mutations also lead to reduced activity of genes surrounding this tumor suppressor gene. Studies have shown that a large number of genes with frequent large fragment deletions on the genome tend to exist in the form of clusters (multiple genes play a common role in the occurrence of cancer), and this common biological regulation is stronger than that of a single gene. The effect is even stronger. Damage on large-scale genomes can act through the co-action of co-occurring cancer genes, rather than through damage to a single independent gene, as a possible mechanism for cancer development.

The process of analyzing candidate tumor suppressor genes is not complicated, and the current analysis protocols are relatively consistent. However, how to obtain common genes from a large number of candidate genes is a difficult point in the current analysis. There are two main methods at present, 1) by using a rat model system of cancer and experimental methods such as RNAi in vivo, this method is time-consuming and expensive. 2) Selecting tumor suppressor genes that have been reported in the literature, due to the limitations of human beings, the screened results are subject to a certain degree, and may be incomplete, and also require more manpower and time.

To sum up, the existing methods of studying tumor suppressor genes will obtain more candidate genes in the early stage, and it is very difficult to pick out the genes that have a single or joint effect on the occurrence of cancer from these genes. According to the method reported in the previous article, it requires a lot of manpower and material resources. Therefore, there is an urgent need in the art for a method for obtaining tumor suppressor genes that play a common role in carcinogenesis from a large number of candidate genes.

SUMMARY OF THE INVENTION

In order to quickly and accurately obtain tumor suppressor genes that play a common role in cancer, the inventors designed a new method:

First, the relationship between all genes in the same region is searched by simulating random sampling;

Second, select genes that have a common relationship;

Then, these genes that co-occur in pairs are linked to form a chain, and the genes in the chain must have a co-occurring relationship;

Finally, combining the clinical information and expression data of the samples, it is verified whether the genes in a chain play a common role in the occurrence of cancer, and whether this common role is stronger than that of a single gene.

The present invention performs downstream analysis on the detection results of somatic CNV (copy number variation) ^[1] and transcriptome expression analysis results (FPKM) ^[1] of tumor samples based on the whole genome composition.

Accordingly, the present invention provides a method for obtaining tumor suppressor genes that act in concert with carcinogenesis, the method comprising the steps of:

1) For tumor tissue samples and normal tissue samples of multiple tumor patients, obtain whole gene sequencing data, transcriptome expression data, and clinical information including patient survival time and gene expression in the samples;

2) dividing the genome into a plurality of sub-regions (for example, the length of the sub-region is 10K-10M, preferably 10K-1M, more preferably 100K-1M), for each sub-region, use the CNV detection results of the above-mentioned whole genome sequencing data calculating CNV significance in the plurality of samples (eg, using a Gscore value ^[2] );

3) Extend the sub-regions with significant changes in CNV (for example, for Gscore value calculation, Gscore>=0.1), and select continuous sub-regions with significant changes in CNV on the genome as deletion regions;

4) For all genes in each deletion region, using the transcriptome expression data of the tumor tissue sample and the normal tissue sample, select the transcriptome expression that has a significant difference between the normal tissue sample and the tumor tissue sample (for example, Paired rank sum test p<0.05) and down-regulated genes, these genes are candidate tumor suppressor genes;

5) For each deletion region, determine whether deletion mutations occur at the same time in two genes, for example, as follows: Assuming that the simultaneous occurrence of deletion mutations in two genes is a random process, then a large number of (for example, more than 10,000 times, 100,000 times) above, ..., more than 10 million times) random sampling, for each sampling, each gene will have cnv in multiple samples, each sampling will target any pair of genes, and the number of times of each sampling is related to the occurrence of the gene. The number of times is related, and the number of times the two genes co-occur (that is, in the same sample) will be recorded for each extraction, and this number of times will be compared with the actual results. The results of the number of times were recorded, the above results were accumulated, and then divided by the total number of samples to obtain the final p value. The smaller the p value, the smaller the randomness of the simultaneous mutation of the two genes. It is generally considered that when p<0.05 When the two genes are considered to have deletion mutations at the same time;

6) Link genes with simultaneous deletion mutations in each region, requiring that the genes in one chain must co-occur with each other;

7) Using the clinical information and the information of the expression level to obtain the difference in the expression of a certain gene during the occurrence of cancer (high expression and low expression in Figure 2), there is a significant difference in the survival rate of cancer ^[3] , analysis The relationship between the expression of genes in the above chain and the survival time of cancer patients, if the genes together have a significant impact on the prognosis of cancer patients (for example, using gene expression data to divide cancer patients into two categories: high and low expression (genes in The expression level in this sample is greater than the mean expression level of the gene in all samples, then the gene is highly expressed in this sample, otherwise it is low expression), calculate the survival rate of each type of patient at a certain time node (survival/ death) as shown in Figure 2, if the high expression of the gene has a higher survival rate for cancer patients, it is verified that the gene has a significant impact on the prognosis of cancer patients), then they do have a common effect, so all Tumor suppressor genes that play a role in carcinogenesis.

Existing methods for studying tumor suppressor genes will obtain more candidate genes in the early stage, and it is very difficult to pick out the genes that really have a single or joint effect on the occurrence of cancer from these genes. According to the method reported in the previous article, it requires a lot of manpower and material resources. The method of this research, clever use of mathematics and bioinformatics methods, can accurately and quickly obtain tumor suppressor genes that act together, and combined with clinical information, it has played a good guide for the follow-up treatment of tumors.

Description of drawings

The final results of the entire study are exemplified by two figures in the accompanying drawings.

Figure 1. Take chromosome 13 as an example to illustrate the genes with large fragment loss mutations on the chromosome, and there is a co-occurring relationship between these genes. This figure shows a large-scale fragment loss on chromosome 13 of the human genome, and the genes associated with this loss mutation form a cluster, and the co-occurrence relationship between them can be shown by the results shown in Figure 2 to verify.

Figure 2. Survival curves of individual genes and their combined effect. It shows that there is indeed a co-occurrence relationship among the clustered genes obtained in this study, and their overall promotion effect on the occurrence of this kind of cancer is obviously greater than that of a single gene. The genes in the figure are also some recently discovered genes that are highly related to the occurrence of cancer in this group.

Detailed ways

This specific embodiment is a further explanation of the present invention, not a limitation of the present invention. After reading this specification, those skilled in the art can make non-creative contribution modifications to the present embodiment according to specific needs, but as long as the claims of the present invention are protected by patent law.

1) Samples and data sources: whole gene sequencing data and transcriptome expression data of tumor tissue samples and normal tissue samples of 65 prostate cancer patients, such as data output and cufflinks using the cnv detection process independently developed by BGI ^{[1] [1]} The FPKM of the gene obtained by the software (the number of fragments aligned per 1K bases of exons in the sequence of the reference genome per 1 million alignments) is input as the data of this example ;

2) Divide each chromosome of the human reference genome (UCSC hg19, http://hgdownload.soe.ucsc.edu/goldenPath/hg19/bigZips/chromFa.tar.gz) into 1M windows, and calculate the CNV in each 1M window Significance (represented by Gscore ^[2] value);

3) Scan the windows obtained in the second step in the order of their positions, and merge adjacent windows with significant CNV changes (Gscore>=0.1) to obtain regions with frequent large-scale losses on the entire genome;

4) Use ANNOVAR (http://www.openbioinformatics.org/annovar/) software to annotate the cnv regions obtained in the third step to obtain a list of genes in these regions;

5) Perform a pairwise rank sum test on the FPKM values of the genes in each region in normal tissue and tumor tissue, and select a list of genes whose expression levels are significantly different (p<0.05) in normal tissue samples and tumor tissue samples ;

6) Filter the resulting gene list: first remove genes with common cancer-related point mutations (somatic single nucleotide mutations and somatic indel mutations) (for example, cancer-related genes in the COSMIC database) List, http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/download), and then select the genes whose expression changes are consistent with the cnv variation, that is, the genes that have been deleted. The expression in the tumor tissue is down-regulated genes;

7) For the genes obtained after screening, according to different cluster regions, study the co-occurrence relationship between the genes in each region: simulate 10,000,000 random sampling, for the results of each sampling, each gene CNV will occur in multiple samples, each sampling will target any pair of genes, the number of times of each sampling is related to the number of occurrences of the gene, and each sampling will record the co-occurrence of these two genes (in the same sample). ), compare this number with the actual result, record the result that the number of simultaneous occurrence of two genes is greater than the actual number of co-occurrence of two genes, accumulate the above results, and then divide by the total number of samples to get The final p value, when p<0.05, considered that the two genes had deletion mutations at the same time;

8) Based on the relationship list obtained in the previous step, connect the genes that co-occur in pairs to obtain a fully interconnected gene list.

9) Draw the cnv region on human chromosome 13, as shown in Figure 1, the abscissa represents the location of chromosome 13 in the human genome, the ordinate represents the Gscore value, the red area represents the amplified area, and the blue area represents the missing Regions, the genes marked in the figure are examples of the gene list found above that co-occur with each other in pairs.

10) Download the clinical information of the published prostate cancer samples (including the patient's survival time) and gene expression in the samples [3], and use the survival curve software package (survival) in the R software to draw each Survival curves of genes and all genes, mfit=survfit(Surv(time, status)～group), where the definition of group: for a single gene, compare the expression value of the gene in the cancer sample with the gene’s expression in the overall sample The mean of the expression in the list is compared, the sample group = 2 (high expression) is greater than the mean, and the sample group = 1 (low expression) is less than the mean, for all genes in the linked list, if all genes are low in this sample Expressed, then group=1 of the entire linked list, otherwise group=2.

The survival curve of the gene in Figure 1 is drawn by the method in the previous step, as shown in Figure 2, the abscissa represents the recurrence time, the ordinate represents the survival probability, the upper curve represents high expression, the lower curve represents low expression, P value It represents whether there is a significant difference in the effect of high and low expression of this gene on the survival rate of patients with prostate cancer (p<0.05 as a significant threshold). As shown in the figure, the tumor suppressor gene obtained by the present invention will have a better prognosis effect when it is highly expressed, and the significant degree of joint action of all genes is greater than that of a single gene.

references

[1] Chiang D Y, Getz G, Jaffe D B, et al. High-resolution mapping of copy-number alterations with massively parallel sequencing[J]. Nature methods, 2008, 6(1): 99-103.

[2] Trapnell C, Roberts A, Goff L, et al.Differential gene and transcriptexpression analysis of RNA-seq experiments with TopHat and Cufflinks[J].Nature protocols,2012,7(3):562-578.

[3] Mermel C H, Schumacher S E, Hill B, et al.GISTIC2.0 facilitatessensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers[J].Genome Biol,2011,12(4):R41 .

[4] Glinsky G V, Glinskii A B, Stephenson A J, et al.Gene expressionprofiling predicts clinical outcome of prostate cancer[J].The Journal ofclinical investigation,2004,113(6):913-923.

[5] Xue W, Kitzing T, Roessler S, et al. A cluster of cooperating tumor-suppressor gene candidates in chromosomal deletions [J]. Proceedings of the National Academy of Sciences, 2012, 109(21): 8212-8217.

[6] J.Clin.Invest.113:913–923(2004).doi:10.1172/JCI200420032.

Claims (6)

1. A method of obtaining a tumor suppressor gene that functions in concert for carcinogenesis, said method comprising the steps of:

1) obtaining whole gene sequencing data, transcriptome expression data, clinical information including patient survival time and gene expression in tumor tissue samples and normal tissue samples of a plurality of tumor patients;

2) dividing a genome into a plurality of subregions, and calculating the CNV significance of the plurality of samples by using the CNV detection result of the whole genome sequencing data for each subregion;

3) extending the sub-regions with significantly changed CNV, and selecting continuous sub-regions with significantly changed CNV on the genome as deletion regions;

4) selecting genes of which the transcriptome expression quantity is obviously different from that of the normal tissue sample and the tumor tissue sample and is down-regulated by using the transcriptome expression quantity data of the tumor tissue sample and the normal tissue sample for all the genes in each deletion region, wherein the genes are candidate tumor suppressor genes;

5) judging whether two genes have deletion mutation or not at the same time in each deletion region;

6) linking the genes which simultaneously have deletion mutation in each region, wherein the genes in one chain must be mutually shared pairwise;

7) obtaining the significant difference in the survival rate of the cancer by the expression difference of a certain gene in the development process of the cancer by using the clinical information and the information of the expression amount, analyzing the relation between the expression of the genes in the chain and the survival time of the cancer patient, and if the genes have significant influence on the prognosis of the cancer patient, the genes have a common effect, so that all tumor suppressor genes which have a common effect on the development of the cancer are obtained;

the length of the subarea is 100K-1M;

the step 3) is specifically as follows: combining adjacent sub-regions with significantly changed CNV to obtain a region frequently lost in a large scale on the whole genome as a deletion region;

step 5) judging whether deletion mutation occurs to every two genes simultaneously is carried out according to the following mode: assuming that the simultaneous occurrence of two gene loss mutations is a random process, random sampling is carried out more than 1000000 ten thousand times, each gene generates cnv in a plurality of samples for each sampling, each sampling aims at any two genes, the frequency of each sampling is related to the frequency of the gene generation, the frequency of the two genes which commonly generate is recorded for each sampling, the frequency is compared with an actual result, the result that the frequency of the two genes which simultaneously generate is greater than the frequency of the two genes which actually generate together is recorded, the results are accumulated, and then the total sampling number is divided to obtain the randomness p of the two genes which simultaneously generate mutations, and if the p is less than 0.05, the two genes simultaneously generate the loss mutations.

2. The method of claim 1, said CNV significance being calculated using a Gscore value.

3. The method of claim 1, said CNV significant change is Gscore > -0.1.

4. The method of claim 1, wherein the transcriptome expression level is significantly different between a normal tissue sample and a tumor tissue sample by a pairwise rank sum test p < 0.05.

5. The method of claim 1, wherein the significant effect of the gene on the prognosis of the cancer patient is confirmed by: the cancer patients are divided into high expression and low expression by using the data of gene expression, the survival rate of each type of patients at a certain time node is calculated, and if the high expression of the gene has higher survival rate to the cancer patients, the gene is verified to have obvious influence on the prognosis of the cancer patients.

6. The method of claim 5, wherein the expression level of the gene in the sample is greater than the mean of the expression levels of the gene in all samples for both high expression and low expression, and the gene is highly expressed in the sample, otherwise the gene is low expressed.

Images