Drug repurposing opportunities across 92 CNS-related conditions using deep learning and whole-genome sequencing

Abstract

CNS-related conditions span tumors, vascular, neurodevelopmental, and psychiatric disorders, yet therapeutic development for CNS disorders continues to face persistent delays and high attrition due to blood–brain barrier constraints, biological heterogeneity, and limited predictive models. Drug repurposing can shorten development timelines but requires scalable, mechanism-grounded prioritization. We curated 213 approved/clinical-stage drugs with KEGG pathway signatures and integrated them with whole-genome sequencing (WGS) pathway-importance profiles from 4392 individuals across 92 diagnoses. For each diagnosis and variant class, the top 30 KEGG pathways were defined as the core signature. Drugs were linked to diagnoses by pathway overlap, excluding on-label indications and infection-only supports, and ranked using a composite Repurposing Score integrating drug maturity/evidence, diagnosis-specific pathway concordance, and WGS-derived genetic support. Score-weight sensitivity analyses increased the contribution of genetic support. Evidence support was assessed by automated screening of publications and ClinicalTrials.gov records. Collectively, 906 drug-linked shared pathways yielded 12,040 unique repurposing pairs; 25.4% were supported by three or more variant classes. Downstream validation-priority annotation identified 1430 high-confidence drug–diagnosis pairs after stratifying candidates by cohort size, variant-class support, supporting-pathway count, and broad KEGG disease-module support. Frequently implicated mechanisms included RTK-MAPK/PI3K, VEGF, immune/checkpoint, and neurotrophin signaling. Variant-class-resolved WGS pathway prioritization coupled to curated pharmacology enables scalable cross-diagnosis repurposing using existing drugs, recovering clinically explored strategies and generating genetically supported hypotheses for biomarker-guided validation.