Modeling SNP Association With Disease Based On Bayesian Information Criterion

Single Nucleotide Polymorphism (SNP) is one of the most common polymorphismin human DNA series, which is caused by single nucleotide variation such as A, T, C, G.It is widely distributed and full of genetic information. As a new genetic marker, SNPmay be relative to many phenotype differences and the infection sensitiveness todiseases or medicine. Now the SNP study is universally accepted as a key step forhuman genome project towards practical applications.This paper is dedicated to modeling SNP association with disease. We proposed amodeling method base on maximum entropy principle and Bayesian informationcriterion. To solve the cross-effect among different disease-caused models, we use asample partition method to classify all the SNP samples, make experiments with SNPsimulation data, work all the probability model out, and compare the result with groundtruth model to verify the correctness of our method. This paper also studies on theapplicability of Bayesian information criterion to modeling SNP disease factors, andproposed an improvement. At last, we have a comparison among AIC, AICC, and BIC inSNP modeling studies, and get a conclusion that BIC is better than the others.