Identification of Potential Molecular Traits Underlying the Genetic Predisposition to Complex diseases Using Multi-omics and Meta-analysis

Abstract

Complex diseases are multifactorial diseases caused by a complex combination of genetic, environmental and lifestyle effects. Numerous non-coding regions that increase the risk for complex diseases have been discovered by successive waves of genome-wide association studies (GWAS). However, the mechanistic understanding underlying GWAS loci has lagged behind GWAS discovery. The rapidly evolving innovations in high throughput molecular profiling technologies have greatly increased our ability to study the downstream transcriptional and epigenetic impacts of disease-associated variants. In my dissertation, I studied the mechanistic underpinning of GWAS loci for complex diseases by using high throughput molecular profiling data and by combining information from multiple studies via meta-analysis.

First, I prioritized mRNAs, microRNAs(miRNAs), and DNA methylation(DNAme) sites potentially involved in Type 2 diabetes (T2D) mechanisms, using data sets in skeletal muscle and subcutaneous adipose tissues from up to 301 individuals from the Finland-United States Investigation of Non-insulin-dependent diabetes mellitus (NIDDM) Genetics (FUSION) Tissue Biopsy Study. I identified quantitative trait loci (QTLs) for mRNAs and miRNAs expression levels and DNAme levels. A smaller proportion of miRNAs had cis-QTLs than mRNAs, and the lead variants for miRNA cis-QTLs had lower minor allele frequency (MAF) than the lead variants for mRNA cis-QTLs. These observations suggest that compared to mRNAs, miRNAs may be under stronger selective pressure and therefore have a lower level of cis-QTL regulation. By integrating the QTLs for molecular traits with T2D GWAS associations, I identified mRNAs and DNAme sites potentially underlying T2D GWAS loci. By testing for associations of molecular trait levels with 48 T2D related traits, we identified mRNAs, miRNAs, and DNAme sites associated with T2D related traits. Multiple lines of evidence suggested that INHBB was likely to underlie the GWAS locus rs11688682 as its eQTL was colocalized with the rs11688682 GWAS locus in both tissues, and INHBB was positively correlated with insulin-related physiological traits in subcutaneous adipose tissue. In addition, the luciferase assay conducted by our collaborators confirmed that the T2D risk allele rs11688682-G increased transcriptional activity in preadipocytes and adipocytes.

Second, I describe a collaborative project using data sets from TwinsUK, METSIM, GTEx and FUSION to perform RNA-seq based eQTL meta-analysis in subcutaneous adipose tissue from 2256 individuals of European ancestry. Of the 19,108 genes present in all studies, the meta-analysis revealed >= 1 eQTL for 15335 (80.3%) genes: 6440 (33.7%) genes had exactly one eQTL, 8895 genes (46.6%) had >= 2 eQTL eQTLs. I evaluated the evidence for colocalization between the meta-analysis eQTLs and the GWAS signals for seven cardiometabolic traits: T2D, Body mass index, Waist-hip ratio, BMI adjusted waist-hip ratio, Coronary artery disease, fasting glucose and fasting insulin. I identified 334 genes that had primary eQTLs colocalized with at least one GWAS signal, and 202 genes that had secondary eQTLs colocalized with at least one GWAS signal.

Throughout my dissertation work, I used molecular profiling data of multiple types of molecular traits and combined eQTL associations from multiple studies to provide clues to the molecular traits that may mediate complex disease risks. These prioritized molecular traits are promising candidates for functional follow-up of their roles in disease etiology.