You are here

    • You are here:
    • Home > Events > PRBB-CRG Sessions Andrew Brown

PRBB-CRG Sessions Andrew Brown

PRBB-CRG Sessions Andrew Brown

31/01/2020
Add to Calendar

PRBB-CRG Sessions Andrew Brown

MARIE CURIE

31/01/202012:00MARIE CURIEPRBB-CRG SessionsAndrew BrownUniversity of Dundee, Scotland, UK"Finding and exploiting causal regulatory variants in the GTEx data"Host: Roderic GuigóAbstract:Fine-mapping of cis-regulatory variants can enhance our understanding of the mechanisms of gene expression and the links between expression and disease. The latest GTEx data provided RNA-seq and WGS from 838 individuals across 49 tissues. We applied 3 published fine-mapping methods to these data: CaVEMaN, CAVIAR and dap-g. We find 68,948 high confidence causal variants (P>0.8), with a median number of 6 variants in a 95% credible set. Fine-mapping approaches can also find independent eQTLs, accounting for multiple testing. We find 7.2% (kidney-cortex) to 45% (nerve-tibial) of eGenes have >1 eQTL, with differences explained by sample size (¿=0.95).

We also used these results to derive properties of regulatory variants; these properties could be used, e.g., for developing personalised genome interpretation methods to predict deleterious variants solely from genomic context. We annotated eQTLs to 715 ChIP-seq peaks and 54 transcription factors motifs, greatest enrichment is in peaks for a mark of active promoters (H3K4me3 peaks contain 16.5% of regulatory variants, a 372 fold enrichment (FE)). Motifs show more enrichment (up to 1,100FE), but explain a far smaller proportion of regulatory variation (0.3%).

We also explored using combinations of annotations to predict causal eQTLs, to produce more tailored inference on specific variants. Assigning priors to variants, 32.2% of variants have unique prior weight due to unique genomic context. Using this information we see a significant increase in the lead variant causal probability (P

We used the fine-mapping results in whole blood to look for colocalization with GWAS hits from 4 lipid traits. We find 36 out of 148 GWAS hits colocalize with eQTLs (P>50%), explaining 21 new GWAS hits compared to the previous GTEx release. This demonstrates the improvement due to the new release, with 332 new samples and genotypes from WGS.

For a majority of eQTLs there remains considerable uncertainty in the precise causal variant. Here we show how leveraging molecular phenotype associations, annotations and posterior probabilities can bring benefits to a diverse set of analyses, from looking at properties of regulatory variants, to assessing allelic heterogeneity, to exploring the relationship with GWAS signals.