Multi-tissue eQTL and pathway analysis of genome-wide genetic association data helps uncover tissue-specific processes of complex disease. A. V. Segrè1, E. R. Gamazon2, D. S. DeLuca1, Y. Meng1, L. D. Ward3, T. Lappalainen4,5, T. Flutre6, X. Wen7, E. T. Dermitzakis4, M. Kellis3, D. L. Nicolae2, N. Cox2, D. G. MacArthur1, K. Ardlie1, G. Getz1, The GTEx Consortium 1) Broad Institute of Harvard and MIT, Cambridge, MA; 2) Department of Medicine, University of Chicago, Chicago, IL; 3) Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, MA; 4) University of Geneva, Department of Genetic Medicine and Development, Genève, Switzerland; 5) Stanford University School of Medicine, Stanford, CA; 6) Department of Human Genetics, University of Chicago, Chicago, IL; 7) Department of Biostatistics, University of Michigan, Ann Harbor, MI.
Genome-wide association studies (GWAS) have identified thousands of common variants (SNPs) associated with complex human diseases, and many more have yet to be found. GWAS interpretation faces multiple challenges, including identifying the causal genes and variants, and understanding their biological mechanisms and tissue-context. Given that pathogenic processes are tissue-specific and a lot of the association signals lie in noncoding regions, integrating genetic variation associated with gene expression changes (eQTLs) from a range of disease-relevant tissues with GWAS SNP data may be invaluable for overcoming these challenges. In support of this, NIH funded the Genotype-Tissue Expression (GTEx) project to generate a large, tissue-wide eQTL resource from ~30 human tissues per individual and hundreds of donors (final goal: 900 donors). To increase statistical and explanatory power, we developed a two-step eQTL-pathway analysis approach, and applied it to eQTLs from 9 tissues in the GTEx pilot project. The first step entails (i) testing whether a set of tissue-specific eQTLs are enriched for multiple modest to strong GWAS associations with a given complex disease or trait compared to a null distribution. If enrichment is found, (ii) genes affected by eQTLs that are top ranked based on their GWAS p-values are tested for enrichment of known biological processes, such as signaling pathways. The statistical framework includes pruning of GWAS SNPs to a set of independent SNPs, mapping eQTLs onto GWAS SNPs using linkage disequilibrium or other co-localization metrics, and evaluating enrichment with permutations. A comprehensive analysis of >10 GWAS and meta-analyses of metabolic, cardiovascular and autoimmune diseases will be presented, using cis-eQTLs from 9 tissues: whole blood, adipose, muscle, heart, artery, lung, skin, nerve and thyroid, taken from 80-185 individuals. Preliminary analysis shows enrichment of multiple modest associations amongst eQTLs in relevant tissues for some diseases, such as whole blood and autoimmune diseases, and less obvious tissue-disease connections for others. Whether eQTLs are the causal mechanisms or the altered function of the affected genes remains to be determined. In addition to proposing key pathogenic tissues for future functional studies, this eQTL-pathway method proposes potential causal genes and biological processes in specific tissues, and provides an initial step for fine-mapping of putative causal eQTL variants.
You may contact the first author (during and after the meeting) at