DrugImproverGPT: A Large Language Model for Drug Optimization with Fine-Tuning via Structured Policy Optimization

• DOI: 10.48550/arxiv.2502.07237
• Publication Date: 2025-02-10
• Authors: Xuefeng Liu; Songhao Jiang; Siyu Chen; Zhiwen Yang; Y. Q. Chen; Ian Foster; Rick Stevens
• Keywords: FOS: Computer and information sciences; Computer Science - Machine Learning; Computer Science - Computation and Language; Quantitative Biology - Biomolecules; Statistics - Machine Learning; FOS: Biological sciences; Biomolecules (q-bio.BM); Machine Learning (stat.ML); Computation and Language (cs.CL); Machine Learning (cs.LG)

Finetuning a Large Language Model (LLM) is crucial for generating results towards specific objectives. This research delves into the realm of drug optimization and introduce novel reinforcement learning algorithm to finetune LLM-based generative model, enhancing original across target objectives, while retains beneficial chemical properties drug. work comprised two primary components: (1) DrugImprover: A framework tailored improving robustness efficiency in optimization. It includes LLM designed Structured Policy Optimization (SPO) algorithm, which theoretically grounded. offers unique perspective fine-tuning model by aligning improvement generated molecule with input under desired (2) dataset 1 million compounds, each OEDOCK docking scores on 5 human proteins associated cancer cells 24 binding sites from SARS-CoV-2 virus. We conduct comprehensive evaluation SPO demonstrate its effectiveness properties. Our code will be publicly available at: https://github.com/xuefeng-cs/DrugImproverGPT.