DrugMCTS: a drug repurposing framework combining multi-agent, RAG and Monte Carlo Tree Search

TL;DR

DrugMCTS is a novel framework that combines retrieval-augmented generation, multi-agent collaboration, and Monte Carlo Tree Search to improve drug repositioning accuracy and robustness beyond traditional language models.

Contribution

It introduces a multi-agent, structured reasoning approach integrating RAG and MCTS, advancing drug repurposing methods with enhanced performance.

Findings

Higher recall and robustness on DrugBank and KIBA datasets

Structured reasoning and agent collaboration improve drug repositioning

Feedback-driven search enhances model effectiveness

Abstract

Recent advances in large language models have demonstrated considerable potential in scientific domains such as drug repositioning. However, their effectiveness remains constrained when reasoning extends beyond the knowledge acquired during pretraining. Conventional approaches, such as fine-tuning or retrieval-augmented generation, face limitations in either imposing high computational overhead or failing to fully exploit structured scientific data. To overcome these challenges, we propose DrugMCTS, a novel framework that synergistically integrates RAG, multi-agent collaboration, and Monte Carlo Tree Search for drug repositioning. The framework employs five specialized agents tasked with retrieving and analyzing molecular and protein information, thereby enabling structured and iterative reasoning. Extensive experiments on the DrugBank and KIBA datasets demonstrate that DrugMCTS achieves substantially higher recall and robustness compared to both general-purpose LLMs and deep learning baselines. Our results highlight the importance of structured reasoning, agent-based collaboration, and feedback-driven search mechanisms in advancing LLM applications for drug repositioning.