Visualizing the Genetic Heterogeneity of Major Depressive Disorder using GDVIS

Investigating the genetic heterogeneity of psychiatric disorders, such as Major Depressive Disorder (MDD) is a key area of research, aimed at increasing etiological insights and ultimately improving clinical outcomes via personalized treatment. Heterogeneity is typically investigated by analyzing disorder subtypes (e.g. depression with/without comorbid anxiety), where heritabilities of the genome-wide association study (GWAS) of each subtype versus controls are estimated, followed by estimation of genetic correlations between these GWASs and of these GWASs with external traits. However, it is hard to synthesize the interconnectedness of these heritability and genetic correlation estimates into a simple overarching understanding of disorder heterogeneity, especially when considering multiple subtype definitions and external traits. Here, we introduce a novel method, Genetic Distance Visualization (GDVIS), which aids interpretation of genetic differences between disorder subtypes, as well as their relation to external traits. GDVIS transforms heritabilities to genetic distances and genetic correlations to angles of a triangular geometric representation of the subtypes. GDVIS requires as input only the two genome-wide association study (GWAS) results of the two subtypes versus controls. All relevant heritabilities (resp. genetic distances) and genetic correlations (resp. angles) are derived based on geometric properties, including the subtype vs. subtype comparison (that was not part of the input). We applied GDVIS to seven subtypes of MDD using data from the UK Biobank. We show that the GDVIS-derived heritabilities and genetic correlations accurately correspond to LDscore-regression estimates. GDVIS illustrates how genetic distances between subtypes can arise from either qualitatively different genetic effects (e.g. for the subtypes of MDD with and without childhood trauma, which only share a limited portion of genetic liability), or from quantitatively different genetic effects (e.g. for the subtypes of MDD with and without recurrence, which completely share genetic risk, but with stronger effects in the subtype with recurrence). Second, we applied GDVIS to relate the MDD subtype-definitions to fifteen external traits. GDVIS visualizations show that subtype vs subtype comparisons can reveal whether the subtypes are differentially associated with the external trait, information that cannot be distilled from the usual subtype vs control comparisons. In conclusion, GDVIS is an effective tool to intuitively and comprehensively visualize the genetic heterogeneity of disorder subtypes.