Exploiting the GTEx resources to decipher the mechanisms at GWAS loci

Abstract

The resources generated by the GTEx consortium offer unprecedented opportunities to advance our understanding of the biology of human diseases. Here, we present an in-depth examination of the phenotypic consequences of transcriptome regulation and a blueprint for the functional interpretation of genome-wide association study-discovered loci. Across a broad set of complex traits and diseases, we demonstrate widespread dose-dependent effects of RNA expression and splicing. We develop a data-driven framework to benchmark methods that prioritize causal genes and find no single approach outperforms the combination of multiple approaches. Using colocalization and association approaches that take into account the observed allelic heterogeneity of gene expression, we propose potential target genes for 47% (2519 out of 5385) of the GWAS loci examined.

Fig. 1

Overview of workflow for mapping complex trait-associated QTLs. Full variant association summary statistics results from 114 GWAS were downloaded, standardized, and imputed to the GTEx v8 WGS variant calls (maf > 0.01) for analyses. A total of 8.87 million imputed and genotyped variants were investigated to identify trait-associated QTLs. A total of 49 tissues, 87 studies (74 distinct traits), and 23,268 protein-coding genes and lncRNAs remained after stringent quality assurance protocols and selection criteria. A wide array of complex trait classes, including cardiometabolic, anthropometric, and psychiatric traits, were included

Fig. 2

Expression and splicing QTL enrichment among GWAS variants. The proportion of genetic variants associated with gene expression (a) and splicing (b) of at least one gene in at least one tissue for each p value cutoff (on x-axis in − log10(p) scale) is shown. The proportions for all tested variants are shown as circles, and the proportions for the GWAS catalog variants are shown as squares

Fig. 3

Dose-dependent effects of QTLs on complex traits. Here, all analyses were performed with fine-mapped variants (QTL with highest posterior inclusion probability). a Schematic representation of dose-response model. b Correlation between QTL and GWAS effects, $\text{Cor}\left(\right|\widehat{\delta }|,|\widehat{\gamma }\left|\right)$. Gray distribution represents permuted null with matched local LD. Each data point corresponds to the median correlation for the trait across 49 tissues. c Average mediated effects from mediation model (${\sigma }_{\text{gene}}^2$, median across tissues). Gray distribution represents permuted null with matched local LD. e Mediated effects of secondary vs. primary eQTLs of genes with colocalization probability (rcp) >0.10. in whole blood, genes for all 87 traits are shown

Fig. 4

Identifying and validating predicted causal genes. a Schematic representation of association and colocalization approaches. b Schematic representation of extrapolating the dose-response curve to the Mendelian end of phenotypic variation spectrum [37]. c Proportion of GWAS-associated loci per trait that contain colocalized and PrediXcan-associated signals for expression and splicing

Fig. 5

Causal gene identification performance. ROC curves of enloc and PrediXcan statistics to identify the “causal” genes (OMIM silver standard) using expression (a) and splicing (b) are shown. Precision recall curves of enloc and PrediXcan to identify silver standard genes using expression (c) and splicing (d) (we show the precision in the range 0 to 0.4 to improve visualization). The number of GWAS loci (LD block-trait pairs) where the OMIM gene was ranked at the top by proximity, enloc, and PrediXcan using expression (e) and splicing (f). In 131 loci out of 206, the OMIM gene was not ranked at the top by either proximity, significance, or colocalization. In thirty one of the loci, the OMIM gene was ranked first by all three criteria. In nineteen loci, the OMIM gene was closest gene (to the top GWAS variant) but not the top gene by PrediXcan significance nor enloc’s colocalization probability

Fig. 6

Identifying trait-relevant tissues using tissue-specific enrichment. Enrichment of tissue-specific association and colocalization compared to the pattern of tissue specificity of non-colocalized genes. Over-representation of the tissue class for PrediXcan-significant and colocalized genes is indicated by dark yellow while depletion is indicated by blue. Black dots label the tissue class-trait pairs passing the nominal p value significance threshold of 0.05. Abbreviation: Table S2. Trait category colors: Fig. S4