An Integrative Computational Approach to Evaluate Genetic Markers for Bipolar Disorder

Abstract

Studies to date have reported hundreds of genes connected to bipolar disorder (BP). However, many studies identifying candidate genes have lacked replication, and their results have, at times, been inconsistent with one another. This paper, therefore, offers a computational workflow that can curate and evaluate BP-related genetic data. Our method integrated large-scale literature data and gene expression data that were acquired from both postmortem human brain regions (BP case/control: 45/50) and peripheral blood mononuclear cells (BP case/control: 193/593). To assess the pathogenic profiles of candidate genes, we conducted Pathway Enrichment, Sub-Network Enrichment, and Gene-Gene Interaction analyses, with 4 metrics proposed and validated for each gene. Our approach developed a scalable BP genetic database (BP_GD), including BP related genes, drugs, pathways, diseases and supporting references. The 4 metrics successfully identified frequently-studied BP genes (e.g. GRIN2A, DRD1, DRD2, HTR2A, CACNA1C, TH, BDNF, SLC6A3, P2RX7, DRD3, and DRD4) and also highlighted several recently reported BP genes (e.g. GRIK5, GRM1 and CACNA1A). The computational biology approach and the BP database developed in this study could contribute to a better understanding of the current stage of BP genetic research and assist further studies in the field.

Figure 1

Diagram for the integrated computational marker evaluation approach.

Figure 2

Validation of different metrics through a LOO cross-validation. (a) Results from GSE35977. (b) Results from GSE82042. Mean CRs of randomly selected genes are displayed in green. The maximum CRs for each metric are presented in corresponding positions.

Figure 3

Top BP genes selected by cross-metrics analysis and their relationships to other diseases. The 11 genes that shared overlap in RScore, PScore and SScore groups are highlighted in red; The 3 genes that shared overlap in AScore, PScore and SScore groups are highlighted in yellow. The network was built using the ‘network building’ module of Pathway Studio.