The Research Of Epistasis Detection Algorithm In Genome-wide Association Study

Complex diseases,also known as common diseases,are the main diseases that afflict mankind.Different from genetically determined Mendelian inheritance disease,causes of complex diseases are complicated,which involve environmental factors,genetic factors and their interactions and many other factors.Among these factors,study on interaction among genes(epistasis)has become an important means to explore etiology of complex diseases.Interactions among genes is mainly embodied in interactions among SNPs,and SNPs is used as genetic marker to test epistasis in genome-wide association study.In recent years,a large number of methods have been proposed to detect epistasis in genome-wide association study,but they still have many problems,such as low efficiency,high false positive rate and not suitable for high-dimensional data sets.To solve the above problems,our study presents NTSACO(A New Two Stage epistasis detection algorithm based on Ant Colony Optimization).The first stage is the stage of screening based on ant colony optimization algorithm,proposed two scoring function to score each feasible solution(SNP locus combinations chosen by ant according to density of pheromone)on the basis of original algorithm,the lower score they get,the closer relations with disease they have,which will transform problem of candidate SNP subset selection to problem of finding optimum solution of two scoring functions,and further transform it to problem of finding non-dominated solutions,and then filtering a non-dominated solution set through a classification algorithm,according to the non-dominated solution concept proposed by Pareto in 1986.The second stage is the detection phase.All selected non-dominated solutions will be exhaustively tested by G~2 test,and eventually returned to the subset of SNP which p-value is lower than significance level.Based on a series of simulated data sets,the experimental results show that the NTSACO algorithm has relatively high epistasis detection efficiency and low false positive rate.And the experimental results on AMD real data sets indicate that NTSACO algorithm is feasible in detecting epistasis.