Date of Award

Fall 12-2018

Degree Name

Doctor of Philosophy (PhD)



Second Advisor


Third Advisor



Metabolomic signatures associated with complex disease have been identified. Metabolomic profiling and the integration of genomic data have proven to be powerful tools to investigate genetic effects underlying intermediate phenotype levels and may facilitate improved understanding of pathophysiologic processes of disease. However, most published studies did not consider sex as an effect modifier, analyze sex-specific effects, nor gene by sex interactions. One reason can be incomplete knowledge of the power of statistical methods used in a given dataset.

I first investigated sex-specific genetic effects by performing sex-stratified exome-wide association studies for 271 chromatography-mass spectrometry measured metabolites in the Atherosclerosis Risk in Communities (ARIC) study, followed by a conventional Z test to evaluate the heterogeneity of genetic effects between men and women. We used African-Americans as the discovery sample and pursued exome-wide significant (false discovery rate Q≤ 5%) genes for replication in European-Americans. Overall, we identified and replicated variants in 12 genes associated with metabolite levels, one of which, rs11555566 in ADA, was a novel common variant suggesting a larger effect in men compared to women for association with N1-methyladenosine levels.

I then focused on rare genetic variants and sex interactions on serum metabolite levels and evaluated the joint effect of genetic main effects and gene-sex interactions in the same discovery and replication population. Using gene-based rareGE and MiSTi approaches, we observed and replicated 14 gene-metabolite associations through joint test, three of which were novel, including PLA2G7- arachidonate (20:4n6), PTER- N-acetyl-beta-alanine and NPC2- leucylserine. Significance of the NPC2- leucylserine association arose from both genetic main effects and gene-sex interaction effects.

Finally yet importantly, I carried out a simulation study to investigate the performance of two aforementioned emerging methods in detecting rare variant gene-sex interaction effects on a quantitative phenotype. Compared with conventional burden tests, rareGE and MiSTi have more power under a wide range of scenarios. Simulation results also illustrate that an approach that jointly tests the genetic main effects and gene-sex interactions increases statistical power and has the potential to uncover novel genetic signals that have not been identified previously.

In conclusion, our study suggests sex-specific genetic effects on the metabolome, and reports novel genetic variants associated with metabolite levels. Use of simulated data provides insights into the power and desired sample size in conducting rare variant G×E interaction studies for these newly introduced methods, justify their use in practice.