Detection, annotation and prioritization of human regulatory variants in the genetics study

Abstract

Interpreting human regulatory variants in the noncoding genomic region is critical to understand the regulatory mechanisms of disease pathogenesis and promote personalized medicine. Recent studies showed that the associated SNPs detected by genome wide association study (GWAS) are significantly enriched in those regions that harbor functional elements, such as transcriptional factor binding sites (TFBSs), chromatin with histone modifications, DNase I hypersensitive sites (DHSs), expression quantitative trait loci (eQTLs) and microRNA (miRNA) binding sites. With the accumulation of functional genomics data, computational methods have been developed to annotate, predict and prioritize noncoding regulatory variants regarding different biological processes. However, evaluating the regulatory effect of genetic variants requires systematic consideration in both transcriptional and post-transcriptional level. In this dissertation, we designed a set of computational methods to predict and prioritize regulatory variants that affect gene regulation with comprehensive evaluations.

We first constructed an integrative database that collect all disease-associated variants from genome wide association study (GWAS). Given the GWAS variants for particular disease/trait, we developed a pipeline GWAS3D to systematically analyze the probability of genetics variants affecting regulatory pathways and underlying disease associations by integrating chromatin state, long range chromosome interaction, sequence motif, and conservation information. We demonstrated that GWAS3D can identify functional regulatory variant that was experimentally validated to affect enhancer function.

Detection and prioritization of regulatory variants in a particular cell/tissue is challenging and requires systematic consideration of chromatin states under corresponding condition. Prediction based on cell type-specific function genomic data can improve the chance and accuracy of regulatory variants discovery. By combining results from multiple methods and epigenome profiles, we developed a Bayesian approach to measure the regulatory potential of genetic variants in a cell type-specific manner. This model can also measure the ensemble effect of chromatin marks around variant locus and estimate regulatory probability of genetic variant on specific cell environment. We showed that this integrative and condition-dependent strategy significantly improves the prediction performance of functional regulatory variants.

Last, we sought to investigate whether genetic variants in the miRNA binding site can affect the function of competing endogenous RNA (ceRNA) and subsequent disease development. Using RNA-seq data on human individuals from different populations, we revealed the genome-wide association between DNA polymorphism and ceRNA regulation. We found regulatory variants can simultaneously affect gene expression changes in both cis and trans through the ceRNA mechanism. We prioritized these variants with their associated ceRNAs according to different criteria and evaluated their collective effect on the ceRNA regulatory network.