MolEvolve: LLM-Guided Evolutionary Search for Interpretable Molecular Optimization

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Chen, Xiangsen, Wu, Ruilong, Lan, Yanyan, Ma, Ting, Liu, Yang
Format:	Preprint
Published:	2026
Subjects:	Machine Learning Artificial Intelligence Computational Engineering, Finance, and Science
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

_version_	1866915890999590912
author	Chen, Xiangsen Wu, Ruilong Lan, Yanyan Ma, Ting Liu, Yang
author_facet	Chen, Xiangsen Wu, Ruilong Lan, Yanyan Ma, Ting Liu, Yang
contents	Despite deep learning's success in chemistry, its impact is hindered by a lack of interpretability and an inability to resolve activity cliffs, where minor structural nuances trigger drastic property shifts. Current representation learning, bound by the similarity principle, often fails to capture these structural-activity discontinuities. To address this, we introduce MolEvolve, an evolutionary framework that reformulates molecular discovery as an autonomous, look-ahead planning problem. Unlike traditional methods that depend on human-engineered features or rigid prior knowledge, MolEvolve leverages a Large Language Model (LLM) to actively explore and evolve a library of executable chemical symbolic operations. By utilizing the LLM to cold start and an Monte Carlo Tree Search (MCTS) engine for test-time planning with external tools (e.g. RDKit), the system self-discovers optimal trajectories autonomously. This process evolves transparent reasoning chains that translate complex structural transformations into actionable, human-readable chemical insights. Experimental results demonstrate that MolEvolve's autonomous search not only evolves transparent, human-readable chemical insights, but also outperforms baselines in both property prediction and molecule optimization tasks.
format	Preprint
id	arxiv_https___arxiv_org_abs_2603_24382
institution	arXiv
publishDate	2026
record_format	arxiv
spellingShingle	MolEvolve: LLM-Guided Evolutionary Search for Interpretable Molecular Optimization Chen, Xiangsen Wu, Ruilong Lan, Yanyan Ma, Ting Liu, Yang Machine Learning Artificial Intelligence Computational Engineering, Finance, and Science Despite deep learning's success in chemistry, its impact is hindered by a lack of interpretability and an inability to resolve activity cliffs, where minor structural nuances trigger drastic property shifts. Current representation learning, bound by the similarity principle, often fails to capture these structural-activity discontinuities. To address this, we introduce MolEvolve, an evolutionary framework that reformulates molecular discovery as an autonomous, look-ahead planning problem. Unlike traditional methods that depend on human-engineered features or rigid prior knowledge, MolEvolve leverages a Large Language Model (LLM) to actively explore and evolve a library of executable chemical symbolic operations. By utilizing the LLM to cold start and an Monte Carlo Tree Search (MCTS) engine for test-time planning with external tools (e.g. RDKit), the system self-discovers optimal trajectories autonomously. This process evolves transparent reasoning chains that translate complex structural transformations into actionable, human-readable chemical insights. Experimental results demonstrate that MolEvolve's autonomous search not only evolves transparent, human-readable chemical insights, but also outperforms baselines in both property prediction and molecule optimization tasks.
title	MolEvolve: LLM-Guided Evolutionary Search for Interpretable Molecular Optimization
topic	Machine Learning Artificial Intelligence Computational Engineering, Finance, and Science
url	https://arxiv.org/abs/2603.24382