A collaborative international study has identified a new set of promising anti-cancer drug candidates using advanced artificial intelligence techniques, with Nigerian scientist Temitope C. Aribigbola contributing to the research.

The study, titled “Deep learning and molecular dynamics reveal promising EZH2 inhibitors for epigenetic cancer targeting,” combined deep learning models, molecular docking, and molecular dynamics simulations to accelerate the search for compounds that can inhibit EZH2 — an epigenetic enzyme linked to several aggressive cancers, including lymphoma, melanoma, breast and prostate cancers.

Aribigbola worked with researchers across multiple institutions, contributing to the methodology and data preparation that supported the computational drug-discovery workflow. The team used a fine-tuned REINVENT generative AI model to design thousands of drug-like molecular structures. More than 8,500 molecules were generated and evaluated for drug-likeness, toxicity, and structural suitability using AI-driven screening.

After several rounds of virtual screening, four compounds — labelled 161, 225, 234 and 383 — showed stronger predicted binding to the EZH2 enzyme than Tazemetostat, the FDA-approved drug currently used in treatment. Some of the molecules recorded binding energies as low as –6.403 kcal/mol, indicating potentially higher effectiveness.

Further molecular dynamics simulations highlighted the stability of these candidates, with compound 383 demonstrating the lowest RMSD and RMSF values, suggesting a strong and consistent interaction with EZH2. Compounds 234 and 225 also showed favourable pharmacokinetic profiles, including good oral absorption and lower predicted cardiotoxicity.

The findings point to the growing role of AI and computational modelling in early-stage drug discovery, an area where researchers like Aribigbola are contributing to ongoing global efforts to develop future cancer therapies.