Summary: A number of integrative statistical methods, including TWAS, allow for gene-level aggregation of association signals from individual genetic variants, leveraging tissue-specific eQTL and GWAS study summary statistics. Genes identified as significant by TWAS may not reflect a causal relationship between gene expression and a downstream trait. Our recently developed statistical method, MRLocus, leverages allelic heterogeneity of cis eQTL signal for a candidate gene, to prioritize genes causally involved in trait variation. Consistent and proportionate effects across multiple cis eSNPs provides evidence beyond existing TWAS approaches that gene expression levels in a particular tissue mediate the heritability of the downstream trait, and allows for estimation of the gene-to-trait effect. We find through various simulation schemes that MRLocus can provide more accurate estimates of these effects as compared to other recently developed cis-MR approaches. Finally, we apply our MRLocus estimation framework to eQTL data from adult post-mortem brain from the PsychENCODE Consortium (n=1,387) and GWAS of schizophrenia (n=40k cases) and bipolar disorder (n=20k cases).
Biography: Dr Michael Love is an Assistant Professor in the Dept. of Biostatistics and Dept. of Genetics at University of North Carolina at Chapel Hill. He obtained his PhD in Computational Biology from the Max Planck Institute for Molecular Genetics and the Freie Universitat in Berlin. The Love Lab uses statistical models to infer biologically meaningful patterns in high-dimensional datasets, and develops open-source statistical software, including for the Bioconductor Project. At UNC-Chapel Hill, we often collaborate with geneticists and epidemiologists, studying how genetic variants relevant to diseases are associated with changes in molecular and cellular phenotypes. Love Lab website: https://mikelove.github.io/ and twitter profile: https://twitter.com/mikelove
All welcome Zoom link