Predicting causal variants affecting expression using whole genome sequence and RNA-seq from multiple human tissues

TY - UNPB

T1 - Predicting causal variants affecting expression using whole genome sequence and RNA-seq from multiple human tissues

AU - Brown, Andrew Anand

AU - Viñuela, Ana

AU - Delaneau, Olivier

AU - Spector, Tim

AU - Small, Kerrin

AU - Dermitzakis, Emmanouil T

PY - 2016/11/21

Y1 - 2016/11/21

N2 - Genetic association mapping produces statistical links between phenotypes and genomic regions, but identifying the causal variants themselves remains difficult. Complete knowledge of all genetic variants, as provided by whole genome sequence (WGS), will help, but is currently financially prohibitive for well powered GWAS studies. To explore the advantages of WGS in a well powered setting, we performed eQTL mapping using WGS and RNA-seq, and showed that the lead eQTL variants called using WGS are more likely to be causal. We derived properties of the causal variant from simulation studies, and used these to propose a method for implicating likely causal SNPs. This method predicts that 25% - 70% of the causal variants lie in open chromatin regions, depending on tissue and experiment. Finally, we identify a set of high confidence causal variants and show that they are more enriched in GWAS associations than other eQTL. Of these, we find 65 associations with GWAS traits and show examples where the gene implicated by expression has been functionally validated as relevant for complex traits.

AB - Genetic association mapping produces statistical links between phenotypes and genomic regions, but identifying the causal variants themselves remains difficult. Complete knowledge of all genetic variants, as provided by whole genome sequence (WGS), will help, but is currently financially prohibitive for well powered GWAS studies. To explore the advantages of WGS in a well powered setting, we performed eQTL mapping using WGS and RNA-seq, and showed that the lead eQTL variants called using WGS are more likely to be causal. We derived properties of the causal variant from simulation studies, and used these to propose a method for implicating likely causal SNPs. This method predicts that 25% - 70% of the causal variants lie in open chromatin regions, depending on tissue and experiment. Finally, we identify a set of high confidence causal variants and show that they are more enriched in GWAS associations than other eQTL. Of these, we find 65 associations with GWAS traits and show examples where the gene implicated by expression has been functionally validated as relevant for complex traits.

U2 - 10.1101/088872

DO - 10.1101/088872

M3 - Preprint

BT - Predicting causal variants affecting expression using whole genome sequence and RNA-seq from multiple human tissues

PB - BioRxiv

ER -