Over 7,000 rare diseases have a characterized genetic mechanism but no approved treatment. Most will never attract the investment required for ground-up drug discovery, yet their molecular biology is often well-documented in public databases. This paper describes a three-module AI pipeline — PPI network analysis, RDKit-based compound filtering, and a referenced multi-criteria scoring function — designed to identify drug repurposing candidates for any rare monogenic disease from public data alone. This is a hypothesis-generating computational study; no wet-lab validation was performed and no clinical inference should be drawn from the results. The pipeline is demonstrated on Charcot-Marie-Tooth disease type 1A (CMT1A), producing a ranked list of 15 compounds across five druggable network targets. Vorinostat (HDAC1, IC50 = 10 nM, LE = 0.577), NMS-873 (VCP, IC50 = 12 nM) and HA15 (HSPA5, IC50 = 320 nM) rank highest. All code is executed; all outputs are reported as generated. Adapting the pipeline to a different disease requires changing only the seed protein.
Andre Henrique Hugendobler (Wed,) studied this question.