Two-stage Multiple Testing Procedures For Genome-wide Gene-environment Interactions

In recent years, gene-environment interaction has caused the people a lot of attention, there are also many methods have been proposed. Two-stage testing is the most prominent, its main idea is to test the interaction of variables in two steps. First, we screen all the genes in the whole genome, then we test the gene-environment interaction in the genes which have been selected in the first stage. In view of the special problem of multiple testing, how to effectively control FWER is a key in resolving the whole problem. In this paper, based on previous studies, different adjustment methods: Bonferroni correction, Sida?k correction, Holm correction for two-stage test have been used in the second step of the existing methods for improvement, which can control FWER well and also make improved efficiency. This paper not only presented some theoretical results of controlling FWER in the two-stage method, but also carried out a feasibility study to further validate the simulation of the proposed method. Finally, our methods were successfully applied to two real data analysis, one is used to test the interaction of the NMRI mouse in the genome-wide, the other one is used to test the interaction of the biochemical index of the patients with II Diabetes mellitus.