PT  - JOURNAL ARTICLE
AU  - Gregory A. Moyerbrailean
AU  - Chris T. Harvey
AU  - Cynthia A. Kalita
AU  - Xiaoquan Wen
AU  - Francesca Luca
AU  - Roger Pique-Regi
TI  - Which genetic variants in DNase I sensitive regions are functional?
AID  - 10.1101/007559
DP  - 2014 Jan 01
TA  - bioRxiv
PG  - 007559
4099  - http://biorxiv.org/content/early/2014/12/05/007559.short
4100  - http://biorxiv.org/content/early/2014/12/05/007559.full
AB  - Ongoing large experimental characterization is crucial to determine all regulatory sequences, yet we do not know which genetic variants in those regions are non-silent. Here, we present a novel analysis integrating sequence and DNase I footprinting data for 653 samples to predict the impact of a sequence change on transcription factor binding for a panel of 1,372 motifs. Most genetic variants in footprints (5,810,227) do not show evidence of allele-specific binding (ASB). In contrast, functional genetic variants predicted by our computational models are highly enriched for ASB (3,217 SNPs at 20% FDR). Comparing silent to functional non-coding genetic variants, the latter are 1.22-fold enriched for GWAS traits, have lower allele frequencies, and affect footprints more distal to promoters or active in fewer tissues. Finally, integration of the annotations into 18 GWAS meta-studies improves identification of likely causal SNPs and transcription factors relevant for complex traits.