Residual genetic connectivity beyond shared liability in 14 psychiatric disorders

Transdiagnostic genetic factor models organise shared liability across psychiatric disorders, but they may leave systematic pairwise genetic overlap unexplained after accounting for shared liability.

Using publicly available Psychiatric Genomics Consortium cross-disorder published genetic correlations, we computed edge-level residual genetic correlations (observed minus model-implied) for all disorder pairs. We summarised residual misfit by ranking the largest residual edges and by aggregating residual edges into disorder-level hub indices. Uncertainty was propagated via Monte Carlo sampling using reported standard errors.

Residual structure was concentrated in a subset of disorders rather than being uniformly distributed. The largest positive residual edge was obsessive-compulsive disorder (OCD)-anxiety (∼0.35), followed by prominent residual links including OCD-Tourette syndrome, attention-deficit/hyperactivity disorder (ADHD)-cannabis use disorder, and autism spectrum disorder (ASD)-ADHD. At the node level, OCD emerged as the most consistent residual hub, with ADHD typically second. Hub rankings did not map one-to-one onto disorder uniqueness, indicating complementary information captured by edge-level residuals.

Higher-order genetic factors capture broad shared liability but leave structured residual associations between specific disorder pairs. Examining residual patterns can clarify the limits of transdiagnostic architectures and inform more explainable models and targeted follow-up.