Linking GWAS to pharmacological treatments for psychiatric disorders

Importance Large-scale genome-wide association studies (GWASs) are expected to inform the development of pharmacological treatments, however the mechanisms of correspondence between the genetic liability identified through GWASs and disease pathophysiology are not well understood.

Objective To investigate whether functional information from a range of open bioinformatics datasets can elucidate the relationship between GWAS-identified genetic variation and the genes targeted by current treatments for psychiatric disorders.

Design, Setting, Participants, and Exposures Relationships between GWAS-identified genetic variation and pharmacological treatment targets were assessed across four psychiatric disorders—ADHD, bipolar disorder, schizophrenia, and major depressive disorder. Using a candidate set of 2232 genes that are listed as targets for all approved treatments in the DrugBank database each gene was independently assigned two scores for each disorder – one based on its involvement as a treatment target, and the other based on the mapping between GWAS-implicated SNPs and genes according to one of four bioinformatic data modalities: SNP position, gene distance on the protein interaction network (PPI), brain eQTL, and gene expression patterns across the brain.

Main Outcomes and Measures Gene scores for pharmacological treatments and GWAS-implicated genes were compared using a novel measure of weighted similarity applying a stringent null hypothesis-testing framework that quantified the specificity of the match by comparing identified associations for a particular disorder to a randomly selected set of treatments.

Results Incorporating information derived from functional bioinformatics data in the form of PPI network revealed links for bipolar disorder (p^perm = 0.0001), however, the overall correspondence between treatment targets and GWAS-implicated genes in psychiatric disorders rarely exceeded null expectations. Exploratory analysis assessing the overlap between the GWAS-identified genetic architecture and treatment targets across disorders identified that most disorder pairs and mapping methods did not show a significant correspondence.

Conclusions and Relevance The relatively low degree of correspondence across modalities suggests that the genetic architecture driving the risk for psychiatric disorders may be distinct from the pathophysiological mechanisms used for targeting symptom manifestations through pharmacological treatments and that novel approaches for understanding and treating psychiatric disorders may be required.