Generative AI-Assisted Drug Repurposing for Oral Cancer

Aim: Accelerate the discovery of existing drugs that can be repurposed to treat oral squamous cell carcinoma using AI-powered virtual screening.

Potential impact: Dramatically reduce drug discovery timeline from 10-15 years to months by identifying promising candidates from 10,000+ approved drugs without expensive lab experiments.

Previous work: We performed transcriptomic profiling across multiple OSCC datasets to identify compounds with inverse transcriptional signatures to OSCC. Using an integrative approach across three drug–gene connectivity platforms, we screened and identified overlapping candidates with potential for repurposing.

Method: To identify drugs against key proteins, the pipeline combines Fpocket-based binding site detection with Structure-based Drug Design using Equivariant Diffusion Models to generate optimal ligand molecules. These are validated via molecular docking and compared to existing drugs using Deep Graph InfoMax embeddings, identifying candidates with the highest binding affinity and structural similarity.

Current progress: The pipeline is deployed and operational. Actively refining molecular embeddings and generation algorithms for improved accuracy.

Next Steps: Laboratory validation using primary cancer cell lines to test top-ranked drug candidates and identify the most effective treatments.