An improved robust association test for GWAS with multiple diseases

In a previous study, we proposed a new design and analysis strategy for Genome-wide Association Studies (GWAS) with multiple diseases but no controls. We have proposed to use an overall chi-square test to test for the association between an SNP with any one of the diseases. The overall chi-square test is not sensitive to the underlying model assumption; however, it does not use the information about the trend among the relative risks of the three genotypes. In this study, we propose a new overall test based on the chi-square partition method. The overall p-value of the proposed approach can be estimated by combining independent p-values from the more powerful one-sided tests which incorporate the trend among the relative risks. Simulation study and real data application show that the proposed test is more powerful and robust.