Orchestrating Tool Ecosystem of Drug Discovery with Intention-Aware LLM Agents

Mingyu Derek Ma, Karina Zadorozhny, Jesse Swanson, Nathan C. Frey, Keunwoo Choi, Maksim Eremeev, Sabrina J Mielke, Wenmo Sun, Melody Liu, Jonathan Wickes, Vladimir Gligorijevic, Richard Bonneau, Henri Dwyer, Kyunghyun Cho, Stephen Ra

Keywords: LLM agents, drug discovery

Abstract:

Fragmented tools and models and complex decision-making with incomplete and heterogeneous information often hinder the drug discovery process. Large Language Models offer promising capabilities in commonsense reasoning and tool integration, yet their application in drug discovery remains constrained by challenges such as being incapable of handling large tool space, limited planning capabilities based on scientific intentions, and unscalable evaluation. We introduce GenieAgent, a drug discovery agent that integrates a wide range of molecule design models and bridges the user intentions to concrete actions by navigating the large skill ecosystem. By unifying disparate tools under a single natural language interface, GenieAgent enables cross-tool reasoning and supports complex scientific workflows. We also propose an evaluation framework simulating drug discovery conversations, based on real-world experiments. A large-scale assessment, validated by expert annotations, demonstrates that GenieAgent reliably meets the majority of molecular engineers' needs with high scientific accuracy and robustness.

Submission Number: 34