Bipolar disorder genetics: Using bioinformatics to guide an investigation of Epistasis.

摘要

This dissertation presents an interrogation of the complex genetic etiology of bipolar disorder (BP). Rather than a single variant resulting in disease, many genetic mechanisms likely contribute to BP's pathophysiology. This work focuses on epistasis, or interactions between genes. Although there is much support for epistasis in BP susceptibility, its identification has been challenging due to lack of adequate genome coverage and sample sizes. The advent of genome-wide association studies has addressed the issue of coverage. The Psychiatric GWAS Consortium provides an unprecedented sample size (11,974 cases and 51,792 controls). Additionally, the field of bioinformatics has made impressive progress in cataloging genetic variants. With the entire genomes of tens of thousands of subjects at our disposal and the ever-expanding annotation of bioinformatics resources to guide our search, we set out to explore epistasis in BP.;The first chapter examined evidence for pairwise epistasis in logistic regression models between ANK3 and genes with bioinformatics support for such interactions. We found intriguing evidence for interactions between ANK3 and voltage-gated potassium channel genes KCNQ2 and KCNQ3. Intrigued by strong biological support for these interactions, in chapter two we reviewed the evidence for how these genes, and ion channels in general, may relate to the BP phenotype. In chapter three, we searched for higher-order interactions using Monte Carlo Logic Regression to prioritize combinations of SNPs for replication with logistic regression. Unfortunately, none of our findings were significant. Finally, in chapter four, we attempted a subphenotype association analysis of suicidality in mood disorder patients within the candidate serotonin pathway, which has been extensively though inconclusively studied. Again, none of our findings survived corrections for multiple testing, but the serotonin receptor gene HTR7 appeared to be suggestively associated across the four different analyses.;This dissertation applies biologically-informed decisions to the investigation of the otherwise intractable undertaking of studying epistasis in high-dimensional datasets. Despite rigorous attempts to methodically interrogate regions of the genome, some of our results were inconclusive while others were null. However, the ANK3-KCNQ2/3 interaction finding is an encouraging result with exciting implications that are worthy of additional investigation.