Probe Before You Edit: Probing-Guided Molecular Optimization for LLM Agents in Structure-Based Drug Design
Abstract
Structure-based drug design increasingly employs LLM agents to iteratively refine ligands against a target pocket, yet a viable ligand must satisfy two often-conflicting objectives---binding affinity and druggability---which single optimization steps rarely improve together. To quantify this difficulty, we introduce two diagnostic metrics: the first measures how often a single edit improves both objectives, and the second measures how often a gain on one objective comes with a loss on the other. Applying these diagnostics to current LLM-agent pipelines exposes a consistent failure mode: the agent performs molecular editing without knowing how the pocket-ligand complex responds to local modifications, thus rarely achieving joint improvement. Inspired by medicinal chemists, who probe the pocket-ligand complex with controlled analog edits before choosing an optimization direction, we propose \textbf{PROBE}, an optimization framework built around edit--response probing. PROBE first decomposes the ligand into editable sites and builds a pocket-specific \textbf{site map} that flags where joint gains are plausible, where the two objectives are likely in tension, and where liability substructures should be changed; it then performs controlled probe edits whose responses are distilled into an \textbf{EditManual}. Guided by the site map and EditManual, PROBE runs an iterative multi-agent optimization loop. On the CrossDocked2020 benchmark, PROBE achieves state-of-the-art performance and substantially mitigates the failure modes exposed by our diagnostics metrics. Codes are in https://github.com/yzf-code/PROBE/tree/main
Related
- Structure-Guided Optimization of a LipL32-Targeting DNA Aptamer through Molecular Dynamics and Free-Energy Analysis
- What Transfers Across Protein Pockets? Experience Reuse in Molecular Optimization for Agentic Structure-Based Drug Design
- Probing the structure of RNAs in solution
- Probing the limits of genetic recoding using multi-omics-guided evolution
- DeCoRE: A computational method to resolve RNA structural heterogeneity from RNA structure probing data by direct RNA sequencing
BibTeX
@misc{yang2026probe,
title = {{Probe Before You Edit: Probing-Guided Molecular Optimization for LLM Agents in Structure-Based Drug Design}},
author = {Zaifei Yang and Weiyu Chen and Yaqing Wang and James Kwok},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2606.00555},
eprint = {2606.00555},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2606.00555},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Probe Before You Edit: Probing-Guided Molecular Optimization for LLM Agents in Structure-Based Drug Design
AU - Zaifei Yang
AU - Weiyu Chen
AU - Yaqing Wang
AU - James Kwok
PY - 2026
DA - 2026/10/23
PB - arXiv
DO - 10.48550/arxiv.2606.00555
UR - https://arxiv.org/abs/2606.00555
AB - Structure-based drug design increasingly employs LLM agents to iteratively refine ligands against a target pocket, yet a viable ligand must satisfy two often-conflicting objectives---binding affinity and druggability---which single optimization steps rarely improve together. To quantify this difficulty, we introduce two diagnostic metrics: the first measures how often a single edit improves both objectives, and the second measures how often a gain on one objective comes with a loss on the other. Applying these diagnostics to current LLM-agent pipelines exposes a consistent failure mode: the agent performs molecular editing without knowing how the pocket-ligand complex responds to local modifications, thus rarely achieving joint improvement. Inspired by medicinal chemists, who probe the pocket-ligand complex with controlled analog edits before choosing an optimization direction, we propose \textbf{PROBE}, an optimization framework built around edit--response probing. PROBE first decomposes the ligand into editable sites and builds a pocket-specific \textbf{site map} that flags where joint gains are plausible, where the two objectives are likely in tension, and where liability substructures should be changed; it then performs controlled probe edits whose responses are distilled into an \textbf{EditManual}. Guided by the site map and EditManual, PROBE runs an iterative multi-agent optimization loop. On the CrossDocked2020 benchmark, PROBE achieves state-of-the-art performance and substantially mitigates the failure modes exposed by our diagnostics metrics. Codes are in https://github.com/yzf-code/PROBE/tree/main
ER -