Nucleotide unit point variation detecting method based on neural network

Abstract

The invention belongs to the technical field of a neural network, and discloses a nucleotide unit point variation detecting method based on a neural network. The method comprises the steps of performing comparison on original fastq data by means of bwa comparing software, and generating a sam file; converting the sam file to a binary-system-form bam file through samtools, performing sequencing onthe bam file and converting to a pileup-format file for finishing preprocessing of the original data; extracting 38 characteristic values which comprise sequencing depth, number of converted bases, base conversion frequency and base; for aiming at the data of the characteristic value, performing training and then storing a training model; detecting the same through the frame of the training modelfor obtaining the SNV. The method can settle a problem of low accuracy in detecting the SNV variation position by means of Fasd-somatic technology, a problem of leakage in detecting the SNV variationby means of somatic snipper method, and a problem of low detecting accuracy on the condition of low tumor purity in prior art.

Description

technical field [0001] The invention belongs to the technical field of neural networks, and in particular relates to a method for detecting single-site variation of nucleotides based on neural networks. Background technique [0002] Currently, the closest available technique: combined sample analysis to detect SNVs. This method is mostly based on the Bayesian model, considering the relationship between normal samples and their paired tumor samples and the purity of the tumor samples, and then setting the threshold to detect SNV, SomaticSniper and Fasd_somatic are both Bayesian models. change. However, the existing schemes also have certain defects. For example, more and more evidences show that the heterogeneity of tumors exists not only within the tumor, but also among tumors; when the sequencing depth is insufficient, some positive mutations with mutational significance are likely to be filtered out. Or it is covered; tumor cells may also have some gene mutations such as...

AI Technical Summary

Problems solved by technology

[0007] 3. The method of detecting SNP, the idea of ​​detecting SNP method: judge whether the site is a SNP by comparing the genotype of a sample with the reference base, and then combine with public databases (such as dbSNPC) to screen out some public mutation sites; The disadvantage of this method is that the undisclosed database SNP will be mistaken for SNV, that is, we can only judge the existing results, but it is not helpful to detect new SNVs, and has little value for cancer research

[0009] (1) Existing methods for detecting SNV mutations on the genome are not accurate enough

[0010] (2) Existing methods for detecting SNV mutations on the genome have low accuracy under low tumor purity

[0011] (3) Existing methods for detecting SNV mutations on the genome rely only on one read on the comparison, which may easily cause the omission of mutation detection

[0013] (1) When the sequencing depth is insufficient, some positive mutations with mutation significance are likely to be filtered or covered

[0014] (2) Tumor cells may also have some gene mutations such as copy number variation and indel

[0015] (3) Due to the presence of non-tumor cells in the tumor body, the depth of the mutation site will be reduced

Images