Analysis of Corrections Methods in Genome-Wide Association Studies

摘要

Genome wide association study (GWAS) can directly study the relationship between human behavior and genotype, which provides a new way for researchers to explore the genetic basis of human behavior from the whole genome level. GWAS involves a large number of sites and behavior of the association test, so we must use multiple correction to control the overall false. Although there are a variety of correction methods to choose from, but the applicability of different correction methods in GWAS research is still lack of systematic research, which makes the choice of multiple correction methods in GWAS lack of theoretical and empirical basis. GWAS commonly used in the correction method is based on the family-wise error rate (FWER) standard Bonferroni correction method, Holm regression adjustment method, permutation test method and false detection rate (FDR) standard BH method. In this paper, the principle and process of the 4 kinds of multiple correction methods are described in detail. A simulation method of GWAS data is presented. The results show that the first 3 methods based on FWER are very small, they are the most stringent control of false, but the number of bits of the real association is significantly lower than the FDR based BH method. Independent data, the BH reported by the SNPs method has the highest rate of interpretation, that is, compared with other methods, the BH method better balance the false and hit. The BH method can be used to correct the results in future studies.