DrugGen-2 integrates disease ontology and protein sequences for molecule design
July 8, 2026
DrugGen-2 generates small molecules by fine-tuning GPT-2 on a dataset of approved drugs, disease links, and protein targets. The model employs supervised fine-tuning followed by reinforcement learning via Group Relative Policy Optimization (GRPO) to optimize for chemical validity and novelty.
HOW THIS AFFECTS YOU
●
researcherThe use of GRPO for molecular property optimization provides a template for multi-objective reinforcement learning in chemistry.
●
healthThis approach shifts molecule generation from target-only to disease-aware contexts.