Statistical Methods To Explore The Phenotypic Consequences Of Gene Regulation

Abstract

BackgroundOver the last 15 years, enormous advances have been made in the technology to interrogate the genome, which in turn has allowed us to discover thousands of genetic variants robustly associated with complex traits including psychiatric disorders. However, most of these variants are not located in coding regions making their interpretation challenging. It is generally accepted that an important mechanism that mediate the genotype phenotype relationship is the regulation of gene expression levels as well as alternative splicing of mRNAs. MethodsWe will describe a method (PrediXcan) to integrate large scale of genome wide association studies (GWAS) with reference transcriptome studies in order to investigate the underlying mechanism behind these discoveries. PrediXcan imputes the transcriptome and computes the correlation between the genetically determined component and the phenotype to prioritize genes that are likely to be causal. ResultsWe will show the results of applying PrediXcan to several large scale GWAS studies using prediction models of 44 human tissues. We find that most associations are tissue specific and that because of sharing between tissues, we increase our chances to discover the mediating genes when we scan across a broad set of tissues. ConclusionsTo take full advantage of the large amounts of genomic data that are being generated, methods that integrate multiple sources of high throughput data are needed. Here we present PrediXcan and its extension that addresses this need. Application to a broad set of phenotypes allows us to explore the phenotypic consequences of gene regulatory variation.