Generative adversarial network-based SNP upper interaction identification method, system and application

Abstract

The invention belongs to the technical field of high-order SNP upper interaction detection, and discloses a generative adversarial network-based SNP upper interaction identification method and system,and application, the adversarial network inputs parameters of an SNP upper model, outputs a sample matrix of an SNP combination, and generates a sample data set according to the parameters of the SNPupper model; the generative adversarial network carries out unified training on the SNP upper model, and the recognizer judges whether an input SNP combination has an upper effect or not and correctly outputs an upper model type to which the upper effect belongs. The invention is different from a traditional machine learning method in which a K-order SNP upper interaction model is learned by using a neural network. The method provided by the invention can be used for quickly identifying the SNP epistatic interaction combination from the whole genome, is high in identification accuracy, has nopreference to the epistatic model, can correctly identify the model types of various different types of SNP epistatic interaction combinations, and provides a theoretical basis for bioscientists.

Description

technical field [0001] The invention belongs to the technical field of high-order SNP epistasis interaction detection, and in particular relates to a SNP epistasis interaction identification method, system and application based on a generative confrontation network. Background technique [0002] At present: SNP (Single Nucleotide Polymorphism, SNP) refers to the polymorphism caused by the variation of a single base site at the genome level, which may be a single base conversion (transition) or transversion (transversion), It may also be due to insertion or deletion of bases. A base pair C-G in sequence 1 is expressed as A-T in sequence 2, and this site is called a SNP site. In the whole human genome, there are more than 3 million such SNP sites. Under normal circumstances, most of the SNPs will not pose a threat to human health, but some SNP variation sites are closely related to human health. SNP epistatic effect: Indicates the interaction between a gene or SNP, tradition...

AI Technical Summary

Problems solved by technology

The main challenges faced by this type of method are: (1) to overcome the problem of multiple testing (2) the threshold is difficult to effectively control, different models, different sample sizes, etc., the statistical thresholds are often different

(3) Cannot identify the type of SNP epistasis

The advantage of this method is that the amount of calculation is small, but it has a preference for SNP epistasis models, and the threshold is difficult to control for SNP combinations of different orders

[0008] At present, for the identification of high-order SNP epistasis combinations, the main problems are: (1) it is difficult to overcome the preference for the SNP epistasis model; (2) the type of SNP epistasis cannot be identified; (3) the traditional method (Bei et al. Yeesian network method, Gini coefficient, mutual information and other methods); (4) Although the method based on machine learning can overcome the recognition preference for the upper model, because it needs to train and test all possible SNP combinations, the calculation The amount is extremely large, and it is difficult to apply to the detection of the whole human genome; (5) Statistical test methods, such as chi-square test, etc., need to carry out multiple tests, the amount of calculation is large, and the statistical threshold is not easy to control

[0010] (1) Existing technologies rely too much on the SNP epistasis model, resulting in a preference for the SNP epistasis model in the identification method, which is difficult to apply to the detection of unknown models

[0011] (2) The P-value threshold used by the existing statistical testing methods is determined artificially, resulting in low recognition accuracy, and it is difficult to control the first type of error and the second type of error

[0012] (3) Existing methods cannot identify the epistasis type to which the SNP epistasis combination belongs

[0014] (1) SNP epistasis interaction models are rich and diverse, and existing methods can only correctly identify a small number of interaction models

[0015] (2) The existing SNP epistasis identification methods are basically statistical calculations based on the genotypes corresponding to the SNP combinations in the sample set, using methods such as multiple testing, and the amount of calculation is very large

[0016] (3) Existing methods can only identify whether a certain SNP combination has an epistasis effect, but cannot determine the type of epistasis effect, and provide insufficient support for later analysis

Images