Scientists at the Southwest Research Institute (SwRI) have developed GAMES (Generative Approaches for Molecular Encodings), a large language model (LLM) designed to enhance drug design and discovery. By generating Simplified Molecular Input Line Entry System (SMILES) strings, GAMES allows researchers to efficiently represent molecular structures, crucial for computational chemistry. Lead developer Dr. Jonathan Bohmann highlights its potential in creating databases for AI processing, facilitating molecular comparisons. Integrated with Rhodium molecular docking software, GAMES optimizes drug discovery, utilizing machine learning with user-friendly SMILES notation. The model was trained on carbon-based molecules and refined through methods like Low-Rank Adaptation (LoRA) and Quantized LoRA (QLoRA), enhancing performance and reducing resource consumption. Research Scientist Daniel Hinojosa emphasizes GAMES’ ability to rank compounds based on “drug-likeness,” vital for regulatory approval. As GAMES advances, SwRI anticipates significant improvements in drug development processes, driving transformative outcomes in the pharmaceutical industry.
Source link
Share
Read more