“Genetical genomics” combines genetic study and gene expression analysis to identify DNA variants that can control gene expression genetically. Such DNA variants are called expression Quantitative Trait Loci (eQTL). Although eQTL studies have great promises, the statistical analysis of eQTL data involving thousands of genes and thousands of markers presents many statistical and computational challenges. In this presentation, we focus one such challenge: multiple testing. In previous studies, the dependency among genes is largely ignored in statistical analysis and consideration of multiple comparison adjustments. However, such dependency may be strong for eQTL data. It can have significant impact on data analysis and interpretation, e.g., false discovery control. We propose to consider both the average as well as the variance of the number of false discoveries in multiple comparison control. This is achieved by developing a variance estimator for the number of false discoveries, and using the upper bound of the false discovery proportion (uFDP) for false discovery control. More importantly, we introduce a weighted version of uFDP (wuFDP) to improve the statistical power to identify eQTLs. The relative performance of uFDP control and wuFDP control in eQTL studies is illustrated through simulation studies and real data analysis.
Forgot your password?
Joel Bader Seminar
Andrew Smith Seminar
Gene Regulatory Networks Special Feature (PNAS)
Abstract