A Novel Two-Stage Approach for Epistasis Detection in Genome-Wide Case-Control Studies

摘要

A significant challenge in epistasis detection is the huge amount of data, which leads to combinatorial explosion. This study focuses on a two-stage approach for detecting epistasis only among single nucleotide polymorphisms (SNPs) that show some marginal effect. We present this two-stage approach based on the fusion of two criteria (TwoFC) to detect epistatic interactions. We fuse the G (2) test and absolute probability difference function as a scoring function to measure the strength of association between SNPs and disease status. The fused scoring function is an excellent measure of the strength of such an association. The two-stage strategy greatly reduces the computation load on epistasis detection. We use both simulated data sets and a real disease data set to evaluate our method. The results of an experiment on the simulated data sets show that TwoFC exhibits high power and sample efficiency. The results of an experiment on the real disease data set show that our method performs well even with large-scale data sets.