CN103218543B - A kind of method and system distinguishing protein coding gene and Noncoding gene - Google Patents

Abstract

The present invention provides a method and system for distinguishing protein-coding genes and non-coding genes, which can distinguish the characteristics of protein-coding genes and non-coding genes at the sequence level, and the characteristics do not depend on the known data of the species and do not require conservation information, and has a good judgment effect on long non-coding RNA. In addition to having a strong advantage in accuracy, it is easy to operate, does not require too much file dependence, and the processing time is significantly better than known methods.

Description

A method and system for distinguishing protein-coding genes from non-coding genes

technical field

The invention relates to the field of life sciences, in particular to a method and system for distinguishing protein-coding genes and non-coding genes.

Background technique

At present, there are mainly two methods in the world to distinguish protein-coding genes (hereinafter referred to as coding genes) and non-coding genes:

The CPC method was developed by the School of Life Sciences of Peking University. It relies on information such as the open reading frame of the predicted gene and the known protein library to determine whether a nucleic acid sequence is a coding gene or a non-coding gene. This method relies too much on the prediction method of the open reading frame and the known database, and there are obvious deficiencies in the judgment of new genes and long non-coding genes. According to our own evaluation, the accuracy of the judgment of long non-coding genes very low.

PhyloCSF is a method adopted internationally in recent years. It relies on sequence comparison information of multiple species to obtain conserved regions, and judges whether it is a coding or non-coding sequence according to the strength of conservation of the sequence to be tested. However, since many species do not have full genome sequences at all, it is not possible to obtain multi-species sequence alignment information. Therefore, for many species, the conservation of the sequence cannot be measured, and thus the coding and non-coding capabilities cannot be determined. In addition, there are multiple conserved modules (subsequences) inside the long non-coding gene, so it is too one-sided to judge the coding ability only by the conserved region, and our own evaluation of this method shows that the accuracy rate is also very low.

Contents of the invention

In order to solve the above problems, the present invention provides a method and system for distinguishing coding genes from non-coding genes, which mainly uses the frequency statistics of sequence tandem codon pairs to accurately classify coding sequences, non-coding sequences and coding regions according to the other five types The sequence generated by the reading method is distinguished, does not depend on the known data of the species, does not require conservative information, and has a good judgment effect on long non-coding RNA.

In order to achieve the above object, the present invention provides a method invention for distinguishing coding genes and non-coding genes, the method comprising:

Step 1, divide the sample set into positive and negative training sets according to the coding and non-coding sequences, and perform steps 2 to 4 for the positive and negative training sets respectively;

Step 2: Count the occurrence frequency of each adjacent nucleotide trimer ANT in the coding sequence, non-coding sequence and intergenic region sequence in the training set and construct the occurrence frequency matrix respectively, based on the three occurrence frequency matrices by The log2-ratio operation constructs the scoring matrix;

Step 3, the scoring matrix calculates the window score S-score by using a sliding window to score, and uses this as the first feature of the classification model, and uses a dynamic programming algorithm to find out the coding sequences in the sample set The area with the maximum subsection sum in the array converted from and non-coding sequences is used as the feature subsequence MLCDS, and the length of the MLCDS is used as the second feature of the classification model;

Step 4, use i∈(1, 2, 3, 4, 5, 6) obtains the third feature of the classification model, where X is the length of the MLCDS in the reading mode, and Yi represents the length of the MLCDS in all six reading modes;

use j∈(1,2,3,4,5) obtains the fourth feature of the classification model, where S is the S-score of MLCDS extracted according to the correct reading method in the six reading methods of the nucleic acid sequence , Ej represents the S-score of MLCDS extracted from the remaining five error code reading methods;

Use the frequency of single nucleotide trimers in coding and non-coding regions to perform log2-ratio calculations to obtain nucleotide trimer preference as the fifth feature of the classification model;

Step 5, using the five features of the positive and negative training sets to form positive and negative feature vector sets to train a classification model, and the sequence to be distinguished is predicted by the classification model to obtain a distinguishing result.

The formula for calculating the frequency of occurrence X _i F is:

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>$

$\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 64</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; 64</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; T</span>}}$

Among them, X represents a certain type of ANT, S _j (X _i ) is the number of occurrences of X in a certain sequence in a certain type of sequence set, and X _i N is the occurrence of this kind of ANT in the entire certain sequence set T represents the total number of occurrences of all types of ANTs in the data set, m represents the number of types of ANTs, and n represents the number of sequences contained in this type of collection.

Further, said step 3 includes:

Step 31, using a sliding window to scan each transcript sequence of the coding sequence and the non-coding sequence according to the six-frame reading code;

Step 32, the scoring matrix will score each sub-window of the sliding window during the scanning process, and convert a transcript composed of nucleotide sequences into six arrays, the elements in the array is the window score of each sub-window;

Step 33, using the method of seeking the maximum subsection sum in the dynamic programming algorithm to find a subsection with the largest sum in each of the six arrays to obtain six candidate maximum fields;

Step 34, the scoring matrix finds the one with the highest score among the six maximum candidate fields as the characteristic subsequence of the transcript most similar to the CDS region.

And the calculation formula of the subsection X with the largest sum in the step 33 is:

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; \}</span>$

a[k] is the largest sub-segment with the largest sub-segment sum, and i and j respectively represent the start and end positions of the largest sub-segment a[k] in this reading mode.

Further, in the step 5: the sequence to be distinguished is divided into positive and negative training sets, then the positive and negative training sets are converted into the required input format of the support vector machine SVM, and put into the SVM Carry out the training of the classification model and obtain the distinguishing result.

To achieve the above object, the present invention also provides a system for distinguishing coding genes and non-coding genes, the system comprising:

The preprocessing module is used to divide the sample set into positive and negative training sets according to the coding and non-coding sequences, and executes the frequency statistics module to the feature extraction module for the positive and negative training sets respectively;

The frequency statistics module is used to count the occurrence frequency of each ANT in the coding sequence, non-coding sequence and intergenic region sequence in the training set and construct the occurrence frequency matrix respectively, which is constructed by log2-ratio operation based on three occurrence frequency matrices scoring matrix;

In the sequence extraction module, the scoring matrix calculates the window score S-score by using a sliding window to score, and uses this as the first feature of the classification model, and uses a dynamic programming algorithm to find out the codes of the sample sequence respectively. Sequence and non-coding sequence are converted into the region with the largest subsection sum in the array as the feature subsequence MLCDS, with the length of the MLCDS as the second feature of the classification model;

The feature extraction module uses i∈(1,2,3,4,5,6) obtains the third feature of the classification model, where X is the length of the MLCDS in the reading mode, and Yi represents the length of the MLCDS in all six reading modes;

use j∈(1,2,3,4,5) obtains the fourth feature of the classification model, where S is the S-score of MLCDS in the reading mode, and Ej represents the S-score of MLCDS in the remaining reading modes. score;

Use the frequency of single nucleotide trimers in coding and non-coding regions to perform log2-ratio calculations to obtain nucleotide trimer preference as the fifth feature of the classification model;

The discrimination result obtaining module uses the five features of the positive and negative training sets to form two sets of positive and negative feature vectors to train the classification model, and the sequence to be distinguished is predicted by the classification model to obtain the discrimination result.

The formula for calculating the frequency of occurrence X _i F is:

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>$

$\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 64</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; 64</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; T</span>}}$

Among them, X represents a certain type of ANT, S _j (X _i ) is the number of occurrences of X in a certain sequence in a certain type of sequence set, and X _i N is the occurrence of this kind of ANT in the entire certain sequence set T represents the total number of occurrences of all types of ANTs in the data set, m represents the number of types of ANTs, and n represents the number of sequences contained in this type of collection.

Further, the sequence extraction module includes:

The scanning module uses a sliding window to scan each transcript sequence of the coding sequence and the non-coding sequence according to the six-frame reading code;

Scoring module, the scoring matrix will score each sub-window of the sliding window during the above-mentioned scanning process, and convert a transcript composed of nucleotide sequences into six arrays, the elements in the array is the window score of each sub-window;

The candidate field acquisition module uses the method of seeking the maximum subsection sum in the dynamic programming algorithm to find a subsection with the largest sum in each of the six arrays to obtain six candidate maximum fields;

In the sequence selection module, the scoring matrix finds the one with the highest score among the six candidate maximum fields as the characteristic subsequence most similar to the CDS region of the transcript.

And the formula for calculating the subsection X with the largest sum in the candidate field acquisition module is:

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; \}</span>$

a[k] is the largest sub-segment with the largest sub-segment sum, and i and j respectively represent the start and end positions of the largest sub-segment a[k] in this reading mode.

Further, in the module for obtaining the distinguishing result: the sequence to be distinguished is divided into positive and negative training sets, and then the positive and negative training sets are converted into the input format required by the support vector machine SVM, and put into Enter the SVM to train the classification model, and obtain the distinction result.

The beneficial effects of the present invention are:

It is an effective method to distinguish coding and non-coding sequences by finding the frequency of tandem codon pairs. The frequency statistics of the tandem codon pairs can accurately generate coding sequences, non-coding sequences and coding regions according to the other five reading methods sequence is distinguished.

Using the idea of dynamic programming, the analysis based on sliding window can extract the subsequence that best represents this sequence from a complete sequence; by using the entire human coding gene transcript as a test set, the subsequences found by our invention More than 98% of the sequences overlap with the standard CDS region in whole or in part.

Differential features of transcripts themselves were proposed. For protein-coding gene transcripts and non-coding gene transcripts, there are not only differences in some attributes and characteristics between the two, but also more significant internal differences between the two. For protein-coding gene transcripts, read according to six frames Translating it into six different sequences, one of them was found to be significantly different from the other five in terms of the length of the representative sequence and the coding ability. But for non-coding gene transcripts this inherent difference does not exist.

The best classification results can be obtained by combining the five characteristics of the maximum field score, length, proportion, value distribution and codon frequency of the sequence.

The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments, but not as a limitation of the present invention.

Description of drawings

Fig. 1a is a flow chart of a method for distinguishing coding genes and non-coding genes of the present invention;

Figure 1b is a schematic diagram of a system for distinguishing coding genes and non-coding genes according to the present invention;

Fig. 2 is the comparative schematic diagram of the ROC curve of the present invention and two kinds of methods of prior art;

Fig. 3 is the occurrence frequency matrix of the present invention;

Figure 4 is a schematic diagram of the tendency of ANT of the present invention to appear on coding and non-coding sequences;

Fig. 5 is a schematic diagram of the S-score distribution of the present invention;

Fig. 6 is a schematic diagram of the sub-window score of the code-reading sequence of the present invention;

Fig. 7 is a schematic diagram of the coincidence degree between the MLCDS of the present invention and the standard CDS sequence;

Fig. 8 is a schematic diagram of statistical comparison between coding and non-coding RNAs of the present invention and average values.

detailed description

Fig. 1a is a flowchart of a method for distinguishing coding genes and non-coding genes according to the present invention. As shown in Figure 1a, the method includes:

Step 1, divide the sample set into positive and negative training sets according to the coding and non-coding sequences, and perform steps 2 to 4 for the positive and negative training sets respectively;

Step 2, count the occurrence frequency of each adjacent nucleotide trimer ANT in the coding sequence, non-coding sequence and intergenic region sequence in the training set and construct the occurrence frequency matrix respectively, based on three occurrence frequency matrices Construct scoring matrix through log2-ratio operation;

Step 3, the scoring matrix calculates the window score S-score by using a sliding window to score, and uses this as the first feature of the classification model, and uses a dynamic programming algorithm to find out the coding sequences in the sample set And the region with the largest sub-segment sum in the array converted into the non-coding sequence is used as the feature sub-sequence MLCDS, and the length of the MLCDS is used as the second feature of the classification model, wherein the largest sub-segment sum refers to a In an array composed of positive and negative numbers, the continuous subsection with the largest sum, that is, if you delete the beginning and end of this continuous subsection or add an adjacent element in the array, the sum of this continuous subsection will be reduced value;

Step 4, use i∈(1,2,3,4,5,6) obtains the third feature of the classification model, where X is the length of the MLCDS in the reading mode, and Yi represents the length of the MLCDS in all six reading modes;

use j∈(1,2,3,4,5) obtains the fourth feature of the classification model, where S is the S-score of MLCDS extracted according to the correct reading method in the six reading methods of the nucleic acid sequence , Ej represents the S-score of MLCDS extracted from the remaining five error code reading methods;

Use the frequency of single nucleotide trimers in coding and non-coding regions to perform log2-ratio calculations to obtain nucleotide trimer preference as the fifth feature of the classification model;

Step 5, using the five features of the positive and negative training sets to form positive and negative feature vector sets to train a classification model, and the sequence to be distinguished is predicted by the classification model to obtain a distinguishing result.

The formula for calculating the frequency of occurrence X _i F is:

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>$

$\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 64</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; 64</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; T</span>}}$

Among them, X represents a certain type of ANT, S _j (X _i ) is the number of occurrences of X in a certain sequence in a certain type of sequence set, and X _i N is the occurrence of this kind of ANT in the entire certain sequence set T represents the total number of occurrences of all types of ANTs in the data set, m represents the number of types of ANTs, and n represents the number of sequences contained in this type of collection.

Further, said step 3 includes:

Step 31, using a sliding window to scan each transcript sequence of the coding sequence and the non-coding sequence according to the six-frame reading code;

Step 32, the scoring matrix will score each sub-window of the sliding window during the scanning process, and convert a transcript composed of nucleotide sequences into six arrays, the elements in the array is the window score of each sub-window;

Step 33, using the method of seeking the maximum subsection sum in the dynamic programming algorithm to find a subsection with the largest sum in each of the six arrays to obtain six candidate maximum fields;

Step 34, the scoring matrix finds the one with the highest score among the six maximum candidate fields as the characteristic subsequence of the transcript most similar to the CDS region.

And the calculation formula of the subsection X with the largest sum in the step 33 is:

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; \}</span>$

a[k] is the largest sub-segment with the largest sub-segment sum, and i and j respectively represent the start and end positions of the largest sub-segment a[k] in this reading mode.

Further, in the step 5: the sequence to be distinguished is divided into positive and negative training sets, then the positive and negative training sets are converted into the required input format of the support vector machine SVM, and put into the SVM Carry out the training of the classification model and obtain the distinguishing result.

Fig. 1b is a schematic diagram of a system for distinguishing coding genes and non-coding genes according to the present invention. As shown in Figure 1b, the system consists of:

The preprocessing module 100 is used to divide the sample set into positive and negative two training sets according to the coding and non-coding sequences, and executes the frequency statistics module to the feature extraction module respectively for the positive and negative two training sets;

The frequency statistics module 200 is used to count the frequency of occurrence of each ANT in the coding sequence, non-coding sequence and intergenic region sequence in the training set and construct the frequency matrix respectively, based on three frequency matrices by log2-ratio operation Build a scoring matrix;

In the sequence extraction module 300, the scoring matrix calculates the window score S-score by using a sliding window to score, and uses this as the first feature of the classification model, and uses a dynamic programming algorithm to find out the S-scores of the sample sequences respectively. The area with the largest sub-segment sum in the array converted from the coding sequence and the non-coding sequence is used as the feature sub-sequence MLCDS, and the length of the MLCDS is used as the second feature of the classification model, wherein the maximum sub-segment sum refers to a In an array composed of positive and negative numbers, the continuous subsection with the largest sum, that is, if you delete the beginning and end of this continuous subsection or add an adjacent element in the array, the sum of this continuous subsection will be reduced value;

The feature extraction module 400 utilizes i∈(1,2,3,4,5,6) obtains the third feature of the classification model, where X is the length of the MLCDS in the reading mode, and Yi represents the length of the MLCDS in all six reading modes;

use j∈(1,2,3,4,5) obtains the fourth feature of the classification model, where S is the S-score of MLCDS in the reading mode, and Ej represents the S-score of MLCDS in the remaining reading modes. score;

Use the frequency of single nucleotide trimers in coding and non-coding regions to perform log2-ratio calculations to obtain nucleotide trimer preference as the fifth feature of the classification model;

The discrimination result obtaining module 500 uses the five features of the positive and negative training sets to form two positive and negative feature vector sets to train the classification model, and the sequence to be distinguished is predicted by the classification model to obtain the discrimination result.

The formula for calculating the frequency of occurrence X _i F is:

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>$

$\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 64</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; 64</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; T</span>}}$

Among them, X represents a certain type of ANT, S _j (X _i ) is the number of occurrences of X in a certain sequence in a certain type of sequence set, and X _i N is the occurrence of this kind of ANT in the entire certain sequence set T represents the total number of occurrences of all types of ANTs in the data set, m represents the number of types of ANTs, and n represents the number of sequences contained in this type of collection.

Further, the sequence extraction module 300 includes:

The scanning module uses a sliding window to scan each transcript sequence of the coding sequence and the non-coding sequence according to the six-frame reading code;

Scoring module, the scoring matrix will score each sub-window of the sliding window during the above-mentioned scanning process, and convert a transcript composed of nucleotide sequences into six arrays, the elements in the array is the window score of each sub-window;

The candidate field acquisition module uses the method of seeking the maximum subsection sum in the dynamic programming algorithm to find a subsection with the largest sum in each of the six arrays to obtain six candidate maximum fields;

In the sequence selection module, the scoring matrix finds the one with the highest score among the six candidate maximum fields as the characteristic subsequence most similar to the CDS region of the transcript.

And the formula for calculating the subsection X with the largest sum in the candidate field acquisition module is:

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; \}</span>$

a[k] is the largest sub-segment with the largest sub-segment sum, and i and j respectively represent the start and end positions of the largest sub-segment a[k] in this reading mode.

Further, in the discrimination result obtaining module 500: the sequence to be distinguished is divided into positive and negative training sets, and then the positive and negative training sets are converted into the input format required by the support vector machine SVM, and its Put it into SVM to train the classification model, and get the distinction result. Middle: Divide the sequence to be distinguished into positive and negative training sets, then convert the positive and negative training sets into the input format required by the support vector machine SVM, and put it into the SVM for classification model training, and obtain Differentiate the results.

Fig. 2 is a schematic diagram comparing the ROC curves of the present invention and the prior art. The present invention uses a test set of mice to compare CNCI, CPC and phyloCSF, and presents a ROC curve. On the basis of the ROC curve, this paper also gives a more intuitive evaluation index: the minimum error rate point, that is, the product of the average false positive rate and false negative rate at this point reaches the lowest, in other words, the minimum error rate The point is the point where the classification model reaches equilibrium, and it is the true performance of the classification model. The entire testing and comparison process is divided into three steps: First, we run CNCI on the local server to calculate the coding and non-coding RNA in the test sample and record the results. Second, we submitted the coding and non-coding RNA data in the test sample to the wed-server of Peking University CPC, and waited for the results to be returned and recorded. Third, in order to run phyloCSF (downloaded from http://compbio.mit.edu/PhyloCSF), we upload the bed file of the test sample and the whole genome annotation information of the mouse to the Galaxy platform, and use the "stitchMAFblocks" function module on this platform , calculate the multiple sequence alignment data of 29 mammals, and convert it into the input format required by phyloCSF under unix, and then run the software with the following parameter configurations according to the help documentation of phyloCSF --minCodons=30, --orf= ATGStop, --strategy=fixed, --frames=3, and --removeRefGaps. These three software will give a score value to the classified sequence, and use zero as the boundary to distinguish coding and non-coding transcripts. The larger the score, the stronger the coding ability of the sequence, otherwise the weaker it is. This article also calculates the ROC curve and determines the minimum error rate by sliding the final scores given by these three softwares. In the figure, a solid line and two different types of dotted lines are used to represent the respective ROC curves of the three softwares, and the minimum error rate point is marked with a solid square In the test data set, CNCI has a more reliable classification effect than CPC and phyloCSF, and lower minimum error rates, which are 0.05, 0.11 and 0.28, respectively (Figure 2). And the time-consuming ability and convenience are more than 3 times better than the other two methods.

The overall structure of CNCI includes two main parts: the construction of CNCI scoring matrix and the establishment of classification model. First, we counted the occurrence frequency of each ANT in the coding sequence, non-coding sequence and intergenic region sequence in the species used as the training set, and constructed three occurrence frequency matrices respectively. Taking the frequency of occurrence of ANT in the intergenic region sequence as a reference, the present invention uses the frequency matrix of ANT in the coding sequence and the frequency matrix of ANT in the non-coding sequence to perform a log2-ratio operation, thereby constructing our CNCI scoring matrix, the matrix It is proposed for the first time in this invention.

Secondly, on the basis of this scoring matrix, we began to build a classification model. At the beginning of the model, the present invention uses a sliding window to scan each transcript sequence according to the six-frame reading code. During the scanning process, the CNCI scoring matrix Each sub-window will be scored, so that a transcript composed of nucleotide sequences is converted into six arrays, and then the method of finding the maximum sub-segment sum in the dynamic programming algorithm is introduced to find out in each array A subsection with the largest sum. After the scan of the six frames, CNCI will automatically find the one with the highest score among the six candidate maximum subsegments as the characteristic subsequence of the transcript, that is, the sequence most similar to the CDS region (MLCDS).

Finally, CNCI calculates the respective MLCDS sets in the coded and non-coded sequence sets of the training samples through the sliding window method, and extracts five features with biological value and statistical significance based on each MLCDS. After that, we send them into the SVM model for training according to the labels of the positive and negative samples. After the training is completed, the work of classifying unknown RNA sequences can be realized.

In this study, we analyzed the frequency of ANT usage in CDS and non-coding regions, which include non-coding RNAs and intronic sequences. There are 64 kinds of codons on the CDS region or nucleotide trimers on the genome sequence, so there are 64×64 kinds of ANTs. Here we count each ANT in three different types of sequences according to the following formula Frequency of occurrence in collection:

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>$

$\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 64</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; 64</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; T</span>}}$

In the above formula, X represents a certain type of ANT, S _j (X _i ) is the number of occurrences of X on a certain sequence in a certain type of sequence set, and X _i N is calculated according to the formula, which means that this type of ANT is in The number of occurrences in the entire set of certain sequences, and T represents the total number of occurrences of all types of ANTs in the data set. Among them, m represents the number of types of ANT, and n represents the number of sequences contained in this type of collection. So X _i F is the frequency of use of a certain ANT in a certain type of data set that we want to calculate.

According to the three data sets of coding region, non-coding RNA and intron sequence, we used the above formula to calculate the occurrence frequency of ANT on the human and mouse RNA data sets, and constructed the occurrence frequency matrix (Figure 3). The abscissa and ordinate respectively represent the types of 64 nucleotide trimers, and the darker the color, the higher the occurrence frequency of this ANT. We can see that the frequency maps of their ANTs in the coding region are very similar for both humans and mice, which shows that humans and mice are still in the same evolutionary model for ANTs, which is also Because the evolutionary distance is not too far apart, the protein-coding regions between the two species are also subject to the same type of selective pressure. Relative to the frequency of use of ANTs in non-coding and intronic sequences, both in humans and mice, there are significant differences from their expression profiles in CDS regions. This set of statistical results shows that ANTs present a biased distribution in coding regions and non-coding regions, and this characteristic also provides a sufficient theoretical basis for us to distinguish coding and non-coding transcripts.

The present invention introduces the frequency of use of ANTs in different gene intervals of humans and mice, and proves that there is a significant difference in the frequency of use of ANTs in the coding region and the non-coding region, but does not give the specific performance of this difference Form, that is, whether a certain ANT tends to appear more in the coding region or in the non-coding region, and whether some ANTs have a specific preference to appear in the coding or non-coding region. We calculated the log2-ratio of the frequency of ANT in the human coding RNA set and the frequency of ANT in the human non-coding RNA set, respectively, and defined the matrix:

$\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; matrix</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; log</span>}<span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; HC</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; matrix</span>}}{\mathrm{<span\; class="notranslate">\; HN</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; matrix</span>}}<span\; class="notranslate">\; )</span>$

The matrix is also defined in the mice collection in the same way:

$\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; matrix</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; log</span>}<span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; MC</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; matrix</span>}}{\mathrm{<span\; class="notranslate">\; MN</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; matrix</span>}}<span\; class="notranslate">\; )</span>$

As shown in Figure 4, each point on it represents the tendency of a certain ANT to appear in the coding and non-coding sequences. The closer the color is to gray, the more the ANT likes to appear in the coding region, and vice versa. The closer to black the more it likes to appear in non-coding areas. It can be seen that the ANTs that are significantly inclined to appear in the non-coding region are all ANTs containing stop codons, that is, these ANTs all start or end with tag, taa and tga and their complementary pairs. The ANTs that tended to appear in coding regions were basically random. This also proves that the frequency of ANT usage is roughly evenly distributed in non-coding regions, while some ANTs are specifically expressed in coding regions. The calculated expression profiles of these two matrices are also basically the same, which fully demonstrates that the scoring matrix is basically universal within the range of evolutionary distance from humans to mice, because human genome data are studied and paid more attention to. There are many, so compared to other species, among the transcript data currently collected by major databases, human data is the largest and the highest quality, so our next training samples are all data from human beings. The mentioned CNCI Classification models also refer to models trained with human RNA data.

According to the CNCI scoring matrix constructed above, we define an index called sequence score (sequence-score: S-score) to measure the coding ability of a sequence. The calculation method of S-score is defined as follows:

$\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span>$

In this calculation expression, S represents the S-score of the sequence, Hp is the CNCI scoring matrix, X represents various types of ANTs, and n represents the length of the sequence after conversion in units of nucleotide trimers . (One nucleotide trimer = 3 nucleotides). After the calculation of the S-score, we can give any sequence a score reflecting its coding ability. The larger the value, the stronger the coding ability of this sequence, and vice versa. The present invention uses the following data to verify whether this S-score can effectively distinguish the real coding sequence, non-coding sequence and the other 5 reading methods of the coding sequence: First, we calculated the six coding methods of 30507 human CDS sequences And the scores of 18566 long non-coding RNAsd, get 7 sets of arrays composed of sequence scores, and divide the sequence scores in these 7 arrays according to the distribution density. Secondly, we apply the human-based CNCI scoring matrix to calculate the true CDS region, the shifted read sequence of the CDS region 1,

The shifted reading frame sequence 2 in the CDS region, the anti-strand reading sequence 1 in the CDS region, the anti-strand reading sequence 2 in the CDS region, the anti-strand reading sequence 3 in the CDS region, and the S-score of the long non-coding RNA (Fig. 5). The black solid line in the figure represents the S-score distribution of the true CDS region, the five black dotted lines represent the S-score distribution of the five misread sequences, and the black shaded part represents the S-score of the code long non-coding RNA Distribution. From the figure, we can clearly see that the real CDS regions generally have higher S-scores, and most of them are greater than zero, while the S-scores of wrong reading methods and long non-coding RNAs are basically the same. distribution below zero, and there is little difference in the distribution of score densities for the two series. According to this, we can get two pieces of information: first, the S-score method can significantly distinguish the coding sequence and the wrong reading method from the non-coding sequence, and second, the wrong reading method and the non-coding sequence are in the S-score There is almost no difference under the division of -score.

The purpose of this study is to develop a method that can distinguish between coding and non-coding RNAs. Although the CNCI scoring matrix can identify CDS regions and other types of non-coding sequences very well, a real coding protein RNA has The 5' non-coding end (5'UTR) and a longer 3' non-coding end (3'UTR), both of which are non-coding regions on the coding protein transcript, their presence will be in a large The S-score of a coding sequence is reduced to a certain extent, so we cannot directly apply this scoring matrix to the full-length RNA sequence. In order to solve this problem, we introduced a sliding window method to assist in the analysis of full-length RNA. The span of this sliding window can be N (10-100) nucleotide trimers, and the window step size is 1 nucleotide Acid trimer, that is, every time you slide, the window will move 3 nucleotides in length. By this method, we can divide a full-length transcript sequence into n pieces (n is the sequence converted to nucleotides The length after the trimer -1, the reason for the conversion is because it is convenient to scan the sequence according to the six reading methods) subsequence, the length of the subsequence is the size of the sliding window. We use the CNCI scoring matrix to score each sub-window and calculate its S-score. Since a full-length transcript sequence has six reading modes, the sliding window will scan any sequence in six ways. times, so under the analysis of the sliding window, any sequence will be converted into 6 independent arrays, and the elements in the array are the S-scores of each sub-window in the original reading method. We used the human coding gene transcript set and the non-coding transcript set in the gencode.v11 database as the test set, calculated the sub-window score distribution of each sequence, and made a comparison of the distribution of the two sets integrate. It can be seen from the map that in the 6 score sets of the same coding gene transcript, we can find a curve with significant difference (indicated by the bold solid line), and the upward protruding part of the curve is the real CDS Regions, but the transcripts of non-coding genes do not have this property. In order to verify the universality and repeatability of this phenomenon found in the present invention, we normalized the 30507 sequences by sampling interpolation method, Scale these sequences to the same length, and the bold solid line represents the sub-window score of the correct reading sequence of all coding genes (Figure 6).

According to the above research, we have found that for coding sequences, there is usually a special reading mode, and there is a continuous and long S-score in the sub-window array of this reading mode, while in non-coding sequences, we do not Did not see this nature. Therefore, the key to solving the RNA classification problem is: how to extract a subsequence with optimal coding ability from coding or non-coding RNA, and ensure that the subsequence from the coding RNA is in the S-score Both the score and the length of the subsequence are significantly higher than those from non-coding subsequences. In this paper, we use the idea of dynamic programming algorithm to extract the above-mentioned subsequence by seeking the maximum subsegment sum. Since this subsequence is basically equivalent to the CDS region in the coding RNA, we named it the most Like a CDS sequence (MLCDS), the entire process of finding the maximum sub-segment sum and finding the MLCDS is as follows:

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; \}</span>$

Here X represents the largest sub-segment sum of a code-reading method, a[k] is the largest sub-segment with this largest sub-segment sum, and i and j represent the largest sub-segment of a[k] in this code-reading mode. The starting and ending positions in , our purpose is to calculate X, i and j, so the following improved formula is defined:

$\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; \{</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\overset{\mathrm{<span\; class="notranslate">\; j</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; \}</span>$

m is a variable, b[j] is the local maximum sub-segment and group, a[j] is the rightmost end of b[j], so we can define the following criteria:

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}}\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span>$

According to the definition of b[j], we can draw a conclusion: when b[j-1]>0, no matter what value a[j] has, b[j]=b[j-1]+a[j] . And when b[j-1]<0, no matter what value a[j] is, b[j]=a[j]. therefore:

$\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; \}</span>$

After the above steps of calculation, we can get the maximum sub-segment sum of the sub-window array of any reading method of any encoding RNA, that is, the subsequence MLCDS with the most coding ability, and the six reading methods It corresponds to six candidate MLCDSs. In this paper, we choose the one with the largest value as the real MLCDS, and the code-reading method it is in is identified as the correct code-reading method. Since we did the same for both coding and non-coding RNAs, one MLCDS would be selected from the non-coding RNAs anyway.

The extraction of MLCDS is the focus of building the entire classification model. It can also be said that the quality of these MLCDS directly determines the final classification effect of CNCI. So in order to prove the credibility of the model, we had to comprehensively evaluate the quality of the MLCDS we found. Since the non-coding RNA does not refer to the CDS region, the present invention refers to 30,507 standard CDSs on hg19, and extracts the reading frames to which these CDS belong and their reading frames according to their original full-length RNA sequences The starting and ending positions of the RNA, and then using this information as a reference condition, compared with the MLCDS we found from the full-length RNA (Figure 7). The various numbers in the figure indicate the degree of coincidence between MLCDS and standard CDS, and the percentages in the pie chart indicate how many percent of MLCDS meet this degree of coincidence. For example, the part represented by the number 1 in the figure shows that in the MLCDS we found , 25% of which have a coincidence degree greater than 95% with the standard CDS sequence. And the number 9 means that only 1.7% of MLCDSs overlap with their corresponding CDSs by less than 40%.

In order to train a classification model to distinguish coding and long non-coding RNAs, five classification features are defined in the present invention. First of all, for coding RNAs, they generally have a long CDS region with high coding quality, while non-coding RNAs do not have this property. The latest research has found that some long non-coding RNAs may contain some relatively high-quality CDS regions. Short peptide chain, but its length and coding ability are far inferior to the real CDS region, and the MLCDS sequence we found can just reflect this characteristic. Therefore, the length of the MLCDS sequence and its S-score value are defined as the two most basic classification features in this paper, and they are named M-Score and M-Length respectively. Secondly, according to our discussion above, in the six reading modes of coding RNA, there will always be a subsequence that is significantly different from the other five ways in the S-score distribution of the sub-window, but non-coding RNA does not. There is no such property. According to this phenomenon, we define two original features from the perspective of the sequence's own differences. This feature extraction method and idea have never been reported in other related literature. The so-called difference in the sequence itself means that if a coding RNA is analyzed, there must be an indicator, which can reflect that there is a special sequence among the six reading methods of this RNA, and this sequence strengthens the However, if it is a non-coding RNA, then this degree of dispersion is relatively insignificant. Therefore, the two new features, LENGTH-Percentage and SCORE-distance are defined as follows:

$\mathrm{<span\; class="notranslate">\; LENGTH</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; Percentage</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; x</span>}}{{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&Element;</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1,2,3,4,5,6</span>}<span\; class="notranslate">\; )</span>$

X is the length of MLCDS in the correct reading mode, and Yi represents the length of MLCDS in all six reading modes.

$\mathrm{<span\; class="notranslate">\; SCORE</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; dis</span>}\mathrm{<span\; class="notranslate">\; the\; tan</span>}\mathrm{<span\; class="notranslate">\; ce</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; E.</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; 5</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&Element;</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1,2,3,4,5</span>}<span\; class="notranslate">\; )</span>$

S is the S-score of MLCDS in the correct reading method, and Ej represents the S-score of MLCDS in the remaining reading methods.

In order to verify that these features can indeed be significantly used to distinguish coding and non-coding sequences, the values calculated for the above four features in their respective types of transcript sets were compared with known coding and non-coding RNAs with good gene annotations. Take a look at the statistics, and find the mean of these four features in the coding and non-coding sets (see Table 1).

Table 1

In order to verify the general applicability of these four features in different species, we selected the two species with the best gene annotations, human and mouse. It can be seen from this that the four features have a good classification effect in these two species, and their calculated values in the coding and non-coding gene sets have obvious differences, so these four features are used by us. Selected to train the classification model. Finally, we defined a 61-dimensional feature (with three stop codons removed) using the frequency of individual nucleotide trimers in the MLCDS region to complement the full CNCI classification model. Nucleotide trimers are called codons in the CDS region. They are key elements that are directly related to the translation of proteins and have biological significance. Since proteins are specific substances that carry biological functions, their components must not be random , there will be no such situation that 20 kinds of amino acids appear uniformly on the polypeptide chain. There must be some codons of proteins necessary for the composition of organisms that are used multiple times in the CDS region. The usage frequency of these trimers or codons must be much higher than the average frequency of 20 amino acids, and not often The opposite happens when used. We still assume that the frequency of occurrence of these trimers in non-coding sequences is approximately equal to the average value, and perform a log2-ratio operation on the frequency of occurrence of these trimers in coding and non-coding regions, so it is more inclined to be in Trimers that occur in coding regions are used more frequently than average, while trimers that tend to occur in non-coding regions are used less frequently. The data set we used for statistics is still 30507 encoded RNAs on hg19 and 18566 non-edited RNAs on Gencode (v11) (Figure 8). The ordinate represents the frequency of the same codon in the coding region relative to the log-ration value of the frequency in the non-coding region, and the abscissa represents the 61 nucleotide trimers except the stop codon. It can be seen that for coding or For non-coding RNA, the vast majority of nucleotide trimers have obvious preferences, and even a few nucleotide trimers with less obvious preferences will play a role in the overall classification effect. Positive weighting effect, so we added this 61-dimensional single nucleotide trimer preference feature to the CNCI model to strengthen the classification effect. So far, all the features of the entire classification model have been defined.

Feature extraction is performed on human coding and non-coding RNA data collections according to the analysis process, and then the obtained feature sequences are divided into positive and negative training sets and organized into the input format required by SVM, and put into it for training of classification models. In this paper, the SVM we use is LIBSVM3.0 version, and the RBF kernel function is selected when training the model, and in order to prevent the occurrence of over-learning phenomenon, all parameters including C and g adopt default values.

Certainly, the present invention also can have other multiple embodiments, without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and deformations according to the present invention, but these corresponding Changes and deformations should belong to the scope of protection of the appended claims of the present invention.

Claims (8)

1. distinguish a method for encoding gene and Noncoding gene, it is characterized in that, comprising:

Step 1, is divided into positive and negative two training set by sample set according to coding and non-coding sequence, performs step 2 respectively to step 4 to positive and negative two training set;

Step 2, counts the frequency of occurrences X of each adjacent nucleotide tripolymer ANT in coded sequence, non-coding sequence and intergenic region sequence in training set _if also builds respectively and occurs frequency matrix, builds scoring matrix, wherein frequency of occurrences X based on three frequency of occurrences matrixes by log2-ratio computing _ithe computing formula of F is:

$X_{i}N=\underset{j=1}{\overset{n}{\&Sigma;}}{S}_j\left(X_i\right)$

$T=\underset{i=1}{\overset{m}{\&Sigma;}}{X}_{i}N=\underset{i=1}{\overset{m}{\&Sigma;}}\left(\underset{j=1}{\overset{n}{\&Sigma;}}{S}_j\right(X_i);m=64*64;n=\left(\mathrm{1........}N\right)$

$X_{i}F=\frac{X_{i}N}{T}$

Wherein X _irepresent the ANT of i-th certain type, S _j(X _i) be occurrence number in a certain bar sequence of X in a certain class arrangement set, X _in is the occurrence number of this kind of ANT in certain arrangement set whole, and T then represents the altogether occurrence number of the ANT of all kinds in certain arrangement set whole, and m represents the kind number of ANT, and n represents the sequence number comprised in the set of the type;

Step 3, the mode that described scoring matrix utilizes moving window to carry out giving a mark calculates window score value S-score, in this, as first feature of disaggregated model, and in the array using dynamic programming algorithm to find out respectively to be converted to by the coded sequence in described sample set and non-coding sequence have maximum subsegment and region as feature subsequence MLCDS, and using the length of described MLCDS as second of disaggregated model feature;

Step 4, utilizes obtain the 3rd feature of disaggregated model, wherein X is the length of MLCDS in reading code mode, and Yi represents MLCDS length respective in whole six kinds of reading code modes;

Utilize obtain the 4th feature of disaggregated model, wherein S is the S-score of the MLCDS extracted according to correct reading code mode in the six kinds of reading code modes had altogether at nucleotide sequence, the S-score of the MLCDS extracted in remaining other the five kinds wrong reading code modes of Ej representative;

Utilize single core thuja acid tripolymer to carry out log2-ratio computing in the frequency of occurrences of coding and non-coding region, obtain five feature of nucleotide tripolymer Preference as disaggregated model;

Step 5, utilizes five positive and negative two incompatible train classification models of set of eigenvectors of feature composition of described positive and negative two training set, treats that distinguishing sequence utilizes described disaggregated model to carry out prediction and obtains distinguishing result.

2. the method distinguishing encoding gene and Noncoding gene as claimed in claim 1, it is characterized in that, described step 2 comprises:

Step 21, uses moving window to scan according to the mode of six frame reading codes every bar transcript sequence of coded sequence and non-coding sequence respectively;

Step 22, described scoring matrix can be given a mark to each subwindow of described moving window in the process of above-mentioned scanning, the transcript that one is made up of nucleotide sequence is converted into six arrays, the element in described array is exactly the window score value of each subwindow;

Step 23, utilize in dynamic programming algorithm ask maximum subsegment and each array in described six arrays of mode in find out one and add and maximum subsegment, obtain six candidate's largest fields;

Step 24, described scoring matrix find out in described six candidate's largest fields maximum that of score value as this transcript as the feature subsequence in CDS region.

3. the method distinguishing encoding gene and Noncoding gene as claimed in claim 2, is characterized in that, adds with maximum subsegment x computing formula to be in described step 23:

$x=\underset{1\&le;i\&le;j\&le;n}{max}\left\{{\&Sigma;}_{k=i}^{j}a\&lsqb;k\&rsqb;\right\}$

A [k] be have this maximum subsegment and maximum subsegment, i and j represents the initial sum final position of this maximum subsegment of a [k] in this reading code mode respectively.

4. the method distinguishing encoding gene and Noncoding gene as claimed in claim 1, it is characterized in that, in described step 5: treat that distinguishing sequence is divided into positive and negative training set by described, then described positive and negative training set is converted to input format required by support vector machines, and put it into the training that SVM carries out disaggregated model, obtain distinguishing result.

5. distinguish a system for encoding gene and Noncoding gene, it is characterized in that, comprising:

Pretreatment module, for sample set is divided into positive and negative two training set according to coding and non-coding sequence, performs frequency statistics module to characteristic extracting module respectively to positive and negative two training set;

Frequency statistics module, for counting the frequency of occurrences X of each ANT in coded sequence, non-coding sequence and intergenic region sequence in training set _if also builds respectively and occurs frequency matrix, builds scoring matrix, wherein frequency of occurrences X based on three frequency of occurrences matrixes by log2-ratio computing _ithe computing formula of F is:

$X_{i}N=\underset{j=1}{\overset{n}{\&Sigma;}}{S}_j\left(X_i\right)$

$T=\underset{i=1}{\overset{m}{\&Sigma;}}{X}_{i}N=\underset{i=1}{\overset{m}{\&Sigma;}}\left(\underset{j=1}{\overset{n}{\&Sigma;}}{S}_j\right(X_i);m=64*64;n=\left(\mathrm{1........}N\right)$

$X_{i}F=\frac{X_{i}N}{T}$

Wherein X _irepresent the ANT of i-th certain type, S _j(X _i) be occurrence number in a certain bar sequence of X in a certain class arrangement set, X _in is the occurrence number of this kind of ANT in certain arrangement set whole, and T then represents the altogether occurrence number of the ANT of all kinds in certain arrangement set whole, and m represents the kind number of ANT, and n represents the sequence number comprised in the set of the type;

Sequential extraction procedures module, the mode that described scoring matrix utilizes moving window to carry out giving a mark calculates window score value S-score, in this, as first feature of disaggregated model, and in the array using dynamic programming algorithm to find out respectively to be converted to by the coded sequence of described sample sequence and non-coding sequence have maximum subsegment and region as feature subsequence MLCDS, using the length of described MLCDS as second of disaggregated model feature;

Characteristic extracting module, utilizes obtain the 3rd feature of disaggregated model, wherein X is the length of MLCDS in reading code mode, and Yi represents MLCDS length respective in whole six kinds of reading code modes;

Utilize obtain the 4th feature of disaggregated model, wherein S is the S-score of MLCDS in the S-score of MLCDS in reading code mode, Ej representative other several reading code modes remaining;

Utilize single core thuja acid tripolymer to carry out log2-ratio computing in the frequency of occurrences of coding and non-coding region, obtain five feature of nucleotide tripolymer Preference as disaggregated model;

Distinguish result and obtain module, utilize five positive and negative two incompatible train classification models of set of eigenvectors of feature composition of described positive and negative two training set, treat that distinguishing sequence utilizes described disaggregated model to carry out prediction and obtains distinguishing result.

6. the system distinguishing encoding gene and Noncoding gene as claimed in claim 5, it is characterized in that, described sequential extraction procedures module comprises:

Scan module, uses moving window to scan according to the mode of six frame reading codes every bar transcript sequence of coded sequence and non-coding sequence respectively;

Scoring modules, described scoring matrix can be given a mark to each subwindow of described moving window in the process of above-mentioned scanning, the transcript that one is made up of nucleotide sequence is converted into six arrays, the element in described array is exactly the window score value of each subwindow;

Candidate's field acquisition module, utilize in dynamic programming algorithm ask maximum subsegment and each array in described six arrays of mode in find out one and add and maximum subsegment, obtain six candidate's largest fields;

Sequence selection module, described scoring matrix find out in described six candidate's largest fields maximum that of score value as this transcript as the feature subsequence in CDS region.

7. the system distinguishing encoding gene and Noncoding gene as claimed in claim 6, is characterized in that, adds with maximum subsegment x computing formula to be in described candidate's field acquisition module:

$x=\underset{1\&le;i\&le;j\&le;n}{max}\left\{{\&Sigma;}_{k=i}^{j}a\&lsqb;k\&rsqb;\right\}$

A [k] be have this maximum subsegment and maximum subsegment, i and j represents the initial sum final position of this maximum subsegment of a [k] in this reading code mode respectively.

8. the system distinguishing encoding gene and Noncoding gene as claimed in claim 5, it is characterized in that, described differentiation result obtains in module: treat that distinguishing sequence is divided into positive and negative training set by described, then described positive and negative training set is converted to input format required by support vector machines, and put it into the training that SVM carries out disaggregated model, obtain distinguishing result.