DDDR-87. Novel drug development for IDH positive gliomas using artificial intelligence prediction models

Abstract Isocitrate dehydrogenase (IDH) enzymes are essential in cellular metabolism and survival by catalyzing the formation of alpha-ketoglutarate (aKG) in the Krebs cycle. Mutations in IDH1 and IDH2 have been established as key contributors to glioma pathogenesis, most evident through their gain-of-function production of D-2-hydroxyglutarate (2HG). This metabolite inhibits aKG-dependent enzymes, promoting cellular dysregulation through DNA modification. Key properties to explore in improving the efficacy of IDH inhibitors include lipophilicity to enhance the ability of compounds to cross cellular membranes, binding adaptability to accommodate variations in target conformation, and stability to maintain efficacy and avoid degradation. AI tools will be imperative to the discovery of more efficacious drugs. Literature was searched to capture three drugs with the highest computer-based predictability of inhibiting alpha-ketoglutarate. A specialized AI-tool modified various structural groups around the three compounds to improve drug interactions along with decreasing the predicted side-effects. Compound 1 modifications included the Addition of a methyl group to cyclopropyl ring. Compound 2 modification included replacing F with CF groups. Compound 3 modification included the addition of a second fluorine atom. Compound 1 showed Increased molecular weight (MW) (241.12 → 255.14), higher LogP (2.73 → 3.12) improving membrane permeability and maintenance of TPSA and H-bonding properties. Compound 2 similarly showed increased MW (296.11 → 310.12), Improved LogP (2.97 → 3.30) maintenance of TPSA and H-bonding profile, stronger electron-withdrawing effect, and better metabolic stability due to CF group. Compound 3 also showed increased MW (296.11 → 314.10), slightly higher LogP (2.97 → 3.11), maintenance of TPSA, enhanced binding through additional halogen bonding, and improved metabolic stability. This study highlights how current AI-tools are able to propose minimal modifications to existing drugs and yet have significantly better predicted effects according to computational predictions. The next step would be to test these drugs on cell lines or IDH-positive glioma organoids to validate their utility.