A Simple Yet Efficient Parametric Method of Local False Discovery Rate Estimation Designed for Genome-Wide Association Data Analysis

In genome-wide association studies, hundreds of thousands of genetic features (genes, proteins, etc.) in a given case-control population are tested to verify existence of an association between each genetic marker and a specific disease. A popular approach in this regard is to estimate local false discovery rate (LFDR), the posterior probability that the null hypothesis is true, given an observed test statistic. However, the existing LFDR estimation methods in the literature are usually complicated. Assuming a chi-square model with one degree of freedom, which covers many situations in genome-wide association studies, we use the method of moments and introduce a simple, fast and efficient approach for LFDR estimation. We perform two different simulation strategies and compare the performance of the proposed approach with three popular LFDR estimation methods. We also examine the practical utility of the proposed method by analyzing a comprehensive 1000 genomes-based genome-wide association data containing approximately 9.4 million single nucleotide polymorphisms, and a microarray data set consisting of genetic expression levels for 6033 genes for prostate cancer patients. The R package implementing the proposed method is available on CRAN <https://cran.r-project.org/web/packages/LFDR.MME>.