Tensor decomposition for multiple-tissue gene expression experiments.

Abstract

Genome wide association studies of gene expression traits and other cellular phenotypes have been successful in revealing links between genetic variation and biological processes. The majority of discoveries have uncovered cis eQTL effects via mass univariate testing of SNPs against gene expression in single tissues. We present a Bayesian method for multi-tissue experiments focusing on uncovering gene networks linked to genetic variation. Our method decomposes the 3D array (or tensor) of gene expression measurements into a set of latent components. We identify sparse gene networks, which can then be tested for association against genetic variation genome-wide. We apply our method to a dataset of 845 individuals from the TwinsUK cohort with gene expression measured via RNA sequencing in adipose, LCLs and skin. We uncover several gene networks with a genetic basis and clear biological and statistical significance. Extensions of this approach will allow integration of multi-omic, environmental and phenotypic datasets.