From shared genetic liability to druggable biology in suicide risk

Suicidal behavior is a clinically heterogeneous and biologically complex outcome. The shared genetic architecture underlying high-risk suicide-related phenotypes and its translational relevance for target discovery remain incompletely defined. Here, we developed an integrative framework combining genomic structural equation modeling (genomic SEM), transcriptome-wide association study (TWAS), causal gene fine‑mapping, artificial intelligence-assisted virtual screening, molecular dynamics simulations, binding free-energy analysis, and integrative causal genomics to characterize a high-suicide-risk syndrome and identify candidate therapeutic targets. Genomic SEM supported a shared latent genetic liability across suicide-related, psychiatric, environmental, socioeconomic, and pain-related phenotypes. Further multidimensional modeling resolved this liability into two biologically interpretable dimensions: endogenous affective risk and exogenous stress. Transcriptomic analyses identified six core risk genes supported by convergent TWAS and fine-mapping evidence: TM9SF4, SUSD3, PTPDC1, VPS33B, XRCC3, and LSM10. Five lead ligands targeting transmembrane 9 superfamily member 4 (TM9SF4) were prioritized following protein-structure prediction and DiffDock-based screening of marine natural products. In the subsequent molecular dynamics simulations and binding free‑energy analyses, all five small molecule–protein complexes maintained stable binding. Integrative analyses of protein and expression quantitative trait loci further identified seven gene associations across neural tissues. Collectively, these findings characterize a genetically informed high-suicide-risk syndrome, indicate TM9SF4 as a potential molecular target, and provide potential candidate natural-product ligands.