Expression quantitative trait loci (eQTL) analyses, which identify geneticmarkers associated with the expression of a gene, are an important tool in theunderstanding of diseases in human and other populations. While most eQTLstudies to date consider the connection between genetic variation andexpression in a single tissue, complex, multi-tissue data sets are now beinggenerated by the GTEx initiative. These data sets have the potential to improvethe findings of single tissue analyses by borrowing strength across tissues,and the potential to elucidate the genotypic basis of differences betweentissues. In this paper we introduce and study a multivariate hierarchical Bayesianmodel (MT-eQTL) for multi-tissue eQTL analysis. MT-eQTL directly models thevector of correlations between expression and genotype across tissues. Itexplicitly captures patterns of variation in the presence or absence of eQTLs,as well as the heterogeneity of effect sizes across tissues. Moreover, themodel is applicable to complex designs in which the set of donors can (i) varyfrom tissue to tissue, and (ii) exhibit incomplete overlap between tissues. TheMT-eQTL model is marginally consistent, in the sense that the model for asubset of tissues can be obtained from the full model via marginalization.Fitting of the MT-eQTL model is carried out via empirical Bayes, using anapproximate EM algorithm. Inferences concerning eQTL detection and theconfiguration of eQTLs across tissues are derived from adaptive thresholding oflocal false discovery rates, and maximum a-posteriori estimation, respectively.We investigate the MT-eQTL model through a simulation study, and rigorouslyestablish the FDR control of the local FDR testing procedure under mildassumptions appropriate for dependent data.
展开▼
