MolEvolve: LLM-Guided Evolutionary Search for Interpretable Molecular Optimization

Xiangsen Chen; Ruilong Wu; Yanyan Lan; Ting Ma; Yang Liu

cs.LGMar 25, 2026

MolEvolve: LLM-Guided Evolutionary Search for Interpretable Molecular Optimization

Authors: Xiangsen Chen, Ruilong Wu, Yanyan Lan, Ting Ma, Yang Liu

Organizations: Hong Kong Polytechnic University · Hong Kong University of Science and Technology(Guangzhou) · Tsinghua University · Harbin Institute of Technology (Shenzhen)

Abstract

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.

Figures & tables

Figure 1: (Up) Existing GNNs act as black boxes with post-hoc interpretations, while LLMs suffer from numerical hallucinations on semantic manifolds. (Down) The representation-precision paradox: minor structural distances lead to ”activity cliffs” in molecular manifolds, which LLMs fail to capture precisely due to semantic smoothing.

Figure 2: The overall architecture of MolEvolve. Phase 1 (Cold Start) distills domain knowledge into executable heuristic rules via symbolic grounding and self-correction. Phase 2 (LLM-MCTS) utilizes these rules to initialize an evolutionary search tree, where the LLM acts as molecule operator to guide selection and expansion within a rigorous verification loop.

Base Model	Method	Regression (RMSE $\downarrow$ )		Classification (ROC-AUC $\uparrow$ )
Base Model	Method	ESOL	Lipo.	BACE	BBBP	HIV
Qwen2.5-7B-Instruct	Automolco ( Zhang et al., 2025 )	1.0566	1.0854	0.6716	0.6102	0.6265
	MolRAG ( Xian et al., 2025 )	1.7160	2.5412	0.6968	0.5608	0.6476
	ATTRILENS-MOL ( Lin et al., 2025 )	1.8420	-	0.6102	0.5810	-
	MolEvolve	0.6890	0.8527	0.7705	0.7461	0.7581
GPT-4o-mini	Automolco ( Zhang et al., 2025 )	1.0062	1.0630	0.7390	0.6435	0.6583

Table 1: Comparative results for molecular property prediction on MoleculeNet benchmarks. For regression tasks (ESOL, Lipophilicity), we report RMSE (lower is better). For classification tasks (BACE, BBBP, HIV), we report ROC-AUC (higher is better).

MolEvolve (Ours)
Base Model	LogP		QED
Base Model	$\Delta$	SR%	$\Delta$	SR%
GPT-4o-mini	1.925	85%	0.264	55%
Qwen2.5-7B-Instruct	1.271	63%	0.245	58%
Qwen2.5-32B-Instruct	2.126	82%	0.258	60%
Baseline Models

Table 2: Molecular property optimization results on ChemCoTBench. Improvement measures the average increase in target property, and SR% (Success Rate) measures the percentage of instances with property improvement. In MolEvolve, the Base Model column specifies the backbone LLM integrated within the architecture.

Figure 3: Search Efficiency and Model Adaptability Analysis: Test RMSE trajectories across 100 iterations. To verify that the evolved symbolic features are model-agnostic, we employ three distinct primary models as downstream evaluators. In each iteration, the MCTS engine generates a candidate feature set, which is then frozen to train these evaluators for scoring.

Figure 4: An Example of Evolutionary Trajectory of Molecular Optimization(logP).

Module	RMSE $\downarrow$		ROC-AUC $\uparrow$
Module	ESOL	Lipo	BACE	BBBP	HIV
Base Model: Qwen2.5-7B-Instruct
Vanilla LLM	1.984	2.251	0.590	0.377	0.621
w/o Cold Start + MCTS	0.831	0.959	0.671	0.691	0.702
w. cold start + MCTS	0.689	0.853	0.771	0.746	0.758
Base Model: GPT-4o-mini

Table 3: Ablation study for molecular property prediction.

Figure 5: Hyperparameter Sensitivity Analysis Heatmaps. We visualize the Average Improvement (Left) and Success Rate (Right) for LogP (Top) and QED (Bottom) tasks. Darker colors indicate superior performance.

Appendix figures & tables5 assets

Supplementary material from the paper’s appendix.

Appendix

Component	Property Prediction (Feature Evolution)	Property Optimization (Structural Evolution)
State ( $s$ )	A tuple of symbolic feature names $\mathcal{F}=(f_{1},f_{2},\dots,f_{k})$ representing a selected descriptor set.	A validated molecular SMILES string $S$ representing a specific chemical structure.
Action ( $a$ )	Selection of a novel RDKit-computable descriptor $f^{\prime}$ from the evolved pool to augment the feature set $\mathcal{F}\rightarrow\mathcal{F}^{\prime}=(f_{1},f_{2},\dots,f_{k},f^{\prime})$ .	A multi-stage structural transformation $S\rightarrow S^{\prime}$ involving LLM-driven critique and rewriting.
Expansion	LLM identifies high-potential descriptors by analyzing physicochemical relevance and existing feature importance.	LLM acts as molecule operator, performing structural diagnosis and modification proposals.
Reward ( $R$ )	Validation metric (e.g., negative RMSE or AUC) augmented by a normalized test signal bonus for exploration stability.	Weighted sum of task score (e.g., LogP/QED), cold-start heuristic scores, and similarity-based penalties.
Verification	Deterministic RDKit-based feature computation; ensures all descriptors in $\mathcal{F}$ are executable.	Enforced by RDKit sanitization and valence checks; invalid SMILES are assigned a heavy penalty ( $R=-100$ ).

Appendix

Table 5: Formal definitions of MCTS components across molecular tasks in our experiments.

Panel of Figure 6 — Figure 6: Real-time Traceable Optimization Case Study. The framework first generates a symbolic diagnosis and prescription (Top). The MCTS engine then executes this plan step-by-step (Bottom). Unlike vanilla LLMs which might hallucinate a jump to the final result, our MolEvolve produces a verifiable edit history where each step corresponds to a specific heuristic rule, providing intrinsic traceable explainability.

Panel of Figure 7 — Figure 7: Step-by-Step MCTS Search History. This detailed view shows how the algorithm explores different branches. Note how Iter 3 (the breakthrough) was achieved by combining the structural simplification from Iter 1 with the high-level symbolic prescription (Nitro replacement), effectively pruning the inferior path taken in Iter 2 .

Panel of Figure 8 — Figure 8: An example of the prompt $\mathcal{P}_{0}$ used in the Knowledge Synthesis phase for ESOL dataset. It serves as the initial instruction to distill chemical intuition into structured textual rules.

Panel of Figure 9 — Figure 9: An example of textual heuristic rules ( $\mathcal{K}_{text}$ ) generated by the LLM for the ESOL prediction task. These rules serve as the semantic basis for the subsequent symbolic code generation.

Molecular Optimization

Molecular optimization computationally designs or edits chemical structures to improve desired properties while respecting chemical and practical constraints.

See all 103 papers

Small Molecule Optimization with Large Language Models
Jul 26, 2024Philipp Guevorguian, Menua Bedrosian, Tigran Fahradyan +3Molecular Optimization Evolutionary Optimization
YerevaNN research lab, Yerevan, Armenia.
My Chemical Harness: Evolutionary Molecular Design over Synthetic Pathways with Large Language Model Agents
Jun 8, 2026César Ojeda, Darius A. Faroughy, Maryam Karimi +3Molecular Optimization LLM Agents
Institute of Mathematics, Faculty of Science, University of Potsdam, Campus Golm, Building 9, Karl-Liebknecht-Str. 24–25, 14476 Potsdam-Golm, Germany · NHETC, Department of Physics and Astronomy, Rutgers University, Piscataway, NJ 08854, USA · Potsdam Transfer, University of Potsdam, Karl-Liebknecht-Str. 24-25, 14476 Potsdam. +1
Deciphering Scientific Reasoning Steps from Outcome Data for Molecule Optimization
Mar 13, 2026Zequn Liu, Kehan Wu, Shufang Xie +5Scientific Reasoning in Language Models Molecular Optimization
Zhongguancun Academy · School of Computer Science and Technology, University of Science and Technology of China · Global Health Drug Discovery Institute

Jul 26, 2024J/K move · Enter open · S save

Small Molecule Optimization with Large Language Models

Philipp Guevorguian, Menua Bedrosian, Tigran Fahradyan, Gayane Chilingaryan+2

YerevaNN research lab, Yerevan, Armenia.

Molecular Optimization Evolutionary Optimization Large Language Model-Guided Optimization AI-Assisted Scientific Research Molecular Generation Drug Discovery

Open paper AskPDF arXiv

Molecular optimization, the process of designing molecules with desirable properties, represents a critical challenge in drug discovery. The recent advancements in large language models (LLMs) have opened new opportunities for their integration with traditional molecular optimization algorithms to improve performance. In this work, we propose Molecular Language Model powered Evolutionary Algorithm (Mol-E), an evolutionary algorithm that relies on the generative capabilities of LLMs trained on molecules and molecular properties. Scientific Contribution. Mol-E obtains the highest aggregate Top-10 AUC among the comparable full-23-task results considered here, scoring 17.500 in the task-agnostic regime, in which the oracle is treated strictly as a black box, and 20.551 in the task-informed regime, in which the optimizer receives a fixed semantic description of the objective. Mol-E also improves over the evaluated baselines on multi-property optimization with docking against DRD2, MK2, and AChE.

Abstract

Molecular Optimization

Explore similar work

Small Molecule Optimization with Large Language Models

My Chemical Harness: Evolutionary Molecular Design over Synthetic Pathways with Large Language Model Agents

Deciphering Scientific Reasoning Steps from Outcome Data for Molecule Optimization