High-resolution Functional Analysis of eQTLs in Multiple Tissues. X. Wen1, R. Pique-Regi2, T. Flutre3, G. Moyerbrailean2, F. Luca2 1) Dept Biostatistics, Univ. of Michigan, Ann Arbor, MI; 2) Center for Molecular Medicine and Genetics, Wayne State University, Detroit, MI; 3) Dept of Human Genetics, University of Chicago, Chicago, IL.
Functional interpretation of eQTLs and their tissue specificity plays a critical role in understanding the transcriptional basis of complex traits. Most existing analyses of this kind are performed within a single tissue type and do not use fine-resolution maps of tissue-specific regulatory sequences. Recently, NIH GTEx and ENCODE projects provide valuable resources to enable the high-resolution functional study of eQTLs across multiple tissues. In this project, we develop novel statistical methods to annotate multiple-tissue eQTLs combining GTEx data from five tissues (blood, skeletal muscle, lung, skin and artery) with ENCODE data from 154 related cell-types.
Our methods hold two distinct advantages over the existing approaches: 1. We extend multiple-tissue eQTL mapping method by Flutre et al 2013 to fine-map potential multiple eQTL signals across multiple tissues simultaneously. This approach not only has high power and specificity in identifying eQTLs and their tissue specificity, but also naturally accounts for LD among candidate SNPs at any given locus. 2. We use tissue-specific single base-pair resolution annotations of genetic variants that are predicted to disrupt transcription factor (TF) binding events. These annotations are obtained by analyzing ENCODE DNase-seq footprints using the CENTIPEDE method (Pique-Regi et al 2011, Degner et al 2012) and a novel extension that learns a new sequence motif model and assesses the potential effect of a genetic variant.
Our preliminary results show that 1. Many genes have multiple cis-eQTLs. We identify examples of eQTLs, for which SNP-by-SNP analysis shows significant opposite effects in a pair of tissues. Through our multiple-SNP multiple-tissue analysis, we are able to convincingly demonstrate that this phenomenon is consistent with the scenario that two distinct tissue-specific eQTLs (active in different tissues) presented and in partial LD. 2. eQTLs are enriched with genetic variants affecting TF bindings. This pattern is consistently observed across examined tissues with high statistical significance (p-value 1.3 x 10-7) 3. The tissue specificity of eQTLs are statistically significantly associated with the tissue-specific TF binding activity in each individual tissue.
You may contact the first author (during and after the meeting) at