Karen Mohlke

At hundreds of complex trait loci identified by genome-wide association studies (GWAS), the underlying noncoding functional variants likely act in cis to alter regulation of nearby genes.  Identification of these variants would aid in understanding the molecular and biological mechanisms by which they influence disease.  To prioritize among sets of candidate variants at GWAS loci, we search for evidence of allele-driven functional variation in sequencing data generated from assays measuring transcription factor binding (ChIP-seq), histone modification (ChIP-seq), and open chromatin (DNase-seq) in disease trait-relevant cell types. To predict specific sites of allelic imbalances indicative of change in gene regulatory activity, we re-map sequence reads from these experiments using an allele-aware analysis pipeline. We then test the predicted variants for allelic differences in transcriptional reporter activity and transcription factor binding in cell-based assays. GWAS loci identified for type 2 diabetes, cholesterol levels, body mass index and waist-hip ratio are significantly enriched for overlap with marks of open chromatin, histone modification, and transcription factor binding in pancreatic islets, liver, brain, and adipose cell types, respectively.  Re-mapping ChIP-seq and DNase-seq reads removes allelic biases, increases the number of peaks detected, and identifies more accurate evidence of allelic imbalances. At several loci associated with type 2 diabetes or HDL-cholesterol levels, we have identified one to four variants that show significant, reproducible allelic differences in transcriptional activity and binding of specific transcription factors.  These *-seq-based predictions and experimental validations provide some of the first molecular mechanisms of regulatory variants at GWAS loci.