大语言模型驱动的分子模拟智能体：从自动化到智能化

韩润一; 徐晓放; 黄剑兴; 李路路; 叶茂; 刘中民

doi:10.11949/0438-1157.20260604

您当前的位置：

首页 >

文章列表页 >

大语言模型驱动的分子模拟智能体：从自动化到智能化

综述与专论 | 更新时间：2026-05-29

- 大语言模型驱动的分子模拟智能体：从自动化到智能化
- Large language model-driven agents for molecular simulation: from automation to intelligence
- 化工学报 2026年
- 作者机构：
  
  1.中国科学院大连化学物理研究所，辽宁大连 116023
  2.中国科学院大学，北京 100049
  3.华为技术有限公司，深圳广东 518129
- 作者简介：
  
  韩润一（2002—），女，博士研究生，hanryi@dicp.ac.cn
  李路路（1992—），男，博士，助理研究员，lilulu@dicp.ac.cn
  叶茂（1973—），男，博士，研究员，maoye@dicp.ac.cn
- 基金信息：
  
  国家自然科学基金项目(22503100);中国科学院项目(XDA0490000)
- DOI：10.11949/0438-1157.20260604
  中图分类号： TQ 021.9
- 收稿：2026-04-29，
  
  修回：2026-05-29，
  
  录用：2026-05-29，
- 稿件说明：
移动端阅览
韩润一, 徐晓放, 黄剑兴, 等. 大语言模型驱动的分子模拟智能体：从自动化到智能化[J/OL]. 化工学报, 2026.

HAN Runyi, XU Xiaofang, HUANG Jianxing, et al. Large language model-driven agents for molecular simulation: from automation to intelligence[J/OL]. CIESC Journal, 2026.
韩润一, 徐晓放, 黄剑兴, 等. 大语言模型驱动的分子模拟智能体：从自动化到智能化[J/OL]. 化工学报, 2026. DOI： 10.11949/0438-1157.20260604.

HAN Runyi, XU Xiaofang, HUANG Jianxing, et al. Large language model-driven agents for molecular simulation: from automation to intelligence[J/OL]. CIESC Journal, 2026. DOI： 10.11949/0438-1157.20260604.

摘要

传统分子模拟方法在处理化工、材料等复杂体系时，常受限于对专家经验的高度依赖、操作流程繁琐与结果可重复性差等瓶颈，难以满足高通量筛选与工业应用的需求。现有的基于脚本的工作流自动化方案虽实现了流程标准化，但其静态预设的执行逻辑无法应对模拟过程中出现的非收敛、力场适配偏差及采样效率低下等动态异常，缺乏自主的科学决策能力。近年来，大语言模型（Large Language Model

LLM）的引入为分子模拟提供了新的逻辑推理引擎与人机交互范式。通过构建集成自然语言指令、领域知识检索增强（Retrieval-augmented Generation

RAG）、模拟流程自动编排及代码自适应生成的智能体，实现了模拟任务从“静态脚本执行”到“动态自适应推演”的转变。本文综述了大语言模型赋能的分子模拟智能体最新进展，重点阐述了其在晶体结构自动构建、力场参数智能推荐、反应路径搜索、自由能计算流程优化及多软件协同仿真中的关键技术路径，分析了智能体在意图理解、复杂任务原子化拆解、全流程工具链编排及异常闭环修正等方面的应用实践，并展望了未来在多模态实验数据融合、基于主动学习的力场自主进化及可信计算框架方面的发展方向，旨在为计算化学与化工过程模拟的智能化升级提供理论参考与技术指引。

Abstract

Traditional molecular simulation methods

when applied to complex systems in chemical engineering and materials science

are often limited by bottlenecks such as a high reliance on expert experience

cumbersome operational procedures

and poor result reproducibility. These limitations hinder their ability to meet the demands of high-throughput screening and industrial applications. Although existing script-based workflow automation schemes have achieved process standardization

their statically preset execution logic fails to address dynamic anomalies during simulations—such as non-convergence

force field mismatches

and low sampling efficiency—thereby lacking autonomous scientific decision-making capabilities. In recent years

the integration of Large Language Models (LLMs) has provided a novel logical reasoning engine and human-computer interaction paradigm for molecular simulation. By constructing agents that integrate natural language instructions

Retrieval-augmented generation (RAG) for domain knowledge

automated orchestration of simulation workflows

and adaptive code generation

a paradigm shift in simulation tasks from "static script execution" to "dynamic adaptive deduction" has been realized. This paper reviews the latest advancements in LLM-empowered molecular simulation agents. It highlights key technical pathways in the automatic construction of crystal structures

intelligent recommendation of force field parameters

reaction pathway searches

optimization of free energy calculation workflows

and collaborative simulations across multiple software platforms. Furthermore

it analyzes the practical applications of these agents in intent understanding

the atomic decomposition of complex tasks

full-process toolchain orchestration

and closed-loop anomaly correction. Finally

this review envisions future development directions

including multimodal experimental data fusion

the autonomous evolution of force fields based on active learning

and trusted computing frameworks

aiming to provide theoretical references and technical guidance for the intelligent upgrade of computational chemistry and chemical process simulations.

关键词

Keywords

references

DeFever R S , Targonski C , Hall S W , et al . A generalized deep learning approach for local structure identification in molecular simulations [J ] . Chemical Science , 2019 , 10 ( 32 ): 7503 - 7515 .

Michel J . Current and emerging opportunities for molecular simulations in structure-based drug design [J ] . Physical Chemistry Chemical Physics , 2014 , 16 ( 10 ): 4465 - 4477 .

De Vivo M , Masetti M , Bottegoni G , et al . Role of molecular dynamics and related methods in drug discovery [J ] . Journal of Medicinal Chemistry , 2016 , 59 ( 9 ): 4035 - 4061 .

Vergadou N , Theodorou D N . Molecular modeling investigations of sorption and diffusion of small molecules in glassy polymers [J ] . Membranes , 2019 , 9 ( 8 ): 98 .

Kee C W . Molecular understanding and practical in silico catalyst design in computational organocatalysis and phase transfer catalysis—Challenges and opportunities [J ] . Molecules , 2023 , 28 ( 4 ): 1715 .

Liu H , Jin Y , Ding H J . MDBuilder: a PyMOL plugin for the preparation of molecular dynamics simulations [J ] . Briefings in Bioinformatics , 2023 , 24 ( 2 ): bbad057 .

Thompson M W , Gilmer J B , Matsumoto R A , et al . Towards molecular simulations that are transparent, reproducible, usable by others, and extensible (TRUE) [J ] . Molecular Physics , 2020 , 118 ( 9/10 ): e1742938 .

Knapp B , Ospina L , Deane C M . Avoiding false positive conclusions in molecular simulation: the importance of replicas [J ] . Journal of Chemical Theory and Computation , 2018 , 14 ( 12 ): 6127 - 6138 .

Hjorth Larsen A , Jørgen Mortensen J , Blomqvist J , et al . The atomic simulation environment—a Python library for working with atoms [J ] . Journal of Physics: Condensed Matter , 2017 , 29 ( 27 ): 273002 .

Horton M . Materials discovery screening with pymatgen [J ] . Nature Reviews Physics , 2023 , 5 ( 2 ): 76 .

Birkenheuer G , Blunk D , Breuers S , et al . MoSGrid–a molecular simulation grid as a new tool in computational chemistry, biology and material science [J ] . Journal of Cheminformatics , 2011 , 3 ( 1 ): P14 .

Di Tommaso P , Chatzou M , Floden E W , et al . Nextflow enables reproducible computational workflows [J ] . Nature Biotechnology , 2017 , 35 ( 4 ): 316 - 319 .

Wei J , Tay Y , Bommasani R , et al . Emergent abilities of large language models [EB/OL ] . 2022 : arXiv : 2206 . 07682 . https://arxiv.org/abs/2206.07682 https://arxiv.org/abs/2206.07682

OpenAI , Achiam J , Adler S , et al . GPT-4 technical report [EB/OL ] . 2023 : arXiv : 2303 . 08774 . https://arxiv.org/abs/2303.08774 https://arxiv.org/abs/2303.08774

Ouyang L , Wu J , Jiang X , et al . Training language models to follow instructions with human feedback [C ] // Advances in Neural Information Processing Systems 35 . 2022 : 27730 - 27744 .

Kumar P . Large language models (LLMs): survey, technical frameworks, and future challenges [J ] . Artificial Intelligence Review , 2024 , 57 ( 10 ): 260 .

Wang X H , Jiang H Y , Yu Y , et al . Building intelligence identification system via large language model watermarking: a survey and beyond [J ] . Artificial Intelligence Review , 2025 , 58 ( 8 ): 249 .

Xiao H G , Zhou F Z , Liu X Y , et al . A comprehensive survey of large language models and multimodal large language models in medicine [J ] . Information Fusion , 2025 , 117 : 102888 .

Yao S Y , Yu D , Zhao J , et al . Tree of thoughts: deliberate problem solving with large language models [C ] // Advances in Neural Information Processing Systems 36 . New Orleans : Neural Information Processing Systems Foundation , 2023 : 11809 - 11822 .

Wei J , Wang X Z , Schuurmans D , et al . Chain-of-thought prompting elicits reasoning in large language models [C ] // Advances in Neural Information Processing Systems 35 . New Orleans : Neural Information Processing Systems Foundation , 2022 : 24824 - 24837 .

Taylor R , Kardas M , Cucurull G , et al . Galactica: a large language model for science [EB/OL ] . 2022 : arXiv : 2211 . 09085 . https://arxiv.org/abs/2211.09085 https://arxiv.org/abs/2211.09085

Wang L , Ma C , Feng X Y , et al . A survey on large language model based autonomous agents [J ] . Frontiers of Computer Science , 2024 , 18 ( 6 ): 186345 .

Boiko D A , MacKnight R , Kline B , et al . Autonomous chemical research with large language models [J ] . Nature , 2023 , 624 ( 7992 ): 570 - 578 .

M Bran A , Cox S , Schilter O , et al . Augmenting large language models with chemistry tools [J ] . Nature Machine Intelligence , 2024 , 6 ( 5 ): 525 - 535 .

Venugopal V , Olivetti E . MatKG: an autonomously generated knowledge graph in Material Science [J ] . Scientific Data , 2024 , 11 ( 1 ): 217 .

Statt M J , Rohr B A , Guevarra D , et al . The materials experiment knowledge graph [J ] . Digital Discovery , 2023 , 2 ( 4 ): 909 - 914 .

Wooldridge M , Jennings N R . Intelligent agents: theory and practice [J ] . The Knowledge Engineering Review , 1995 , 10 ( 2 ): 115 - 152 .

Xi Z H , Chen W X , Guo X , et al . The rise and potential of large language model based agents: a survey [J ] . Science China Information Sciences , 2025 , 68 ( 2 ): 121101 .

Qin Y J , Liang S H , Ye Y N , et al . ToolLLM: Facilitating large language models to master 16000+ real-world APIs [C ] // International Conference on Learning Representations . 2024 : 9695 - 9717 .

Shinn N , Cassano F , Gopinath A , et al . Reflexion: Language agents with verbal reinforcement learning [C ] // Advances in Neural Information Processing Systems 36 . New Orleans : Neural Information Processing Systems Foundation , 2023 : 8634 - 8652 .

Sumers T R , Yao S Y , Narasimhan K , et al . Cognitive architectures for language agents [J ] . Transactions on Machine Learning Research , 2023 .

Durante Z , Huang Q , Wake N , et al . Agent AI: surveying the horizons of multimodal interaction [EB/OL ] . 2024 : arXiv : 2401 . 03568 . https://arxiv.org/abs/2401.03568 https://arxiv.org/abs/2401.03568

Wang R Y , Jansen P , Côté M A , et al . ScienceWorld: is your agent smarter than a 5th grader? [C ] // Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing . Abu Dhabi : Association for Computational Linguistics , 2022 : 11279 - 11298 .

Shen Y L , Song K T , Tan X , et al . HuggingGPT: solving AI tasks with ChatGPT and its friends in Hugging Face [C ] // Advances in Neural Information Processing Systems 36 . New Orleans : Neural Information Processing Systems Foundation , 2023 : 38154 - 38180 .

Yang Z Y , Li L J , Wang J F , et al . MM-REACT: prompting ChatGPT for multimodal reasoning and action [EB/OL ] . 2023 : arXiv : 2303 . 11381 . https://arxiv.org/abs/2303.11381 https://arxiv.org/abs/2303.11381

Huang R J , Li M Z , Yang D C , et al . AudioGPT: understanding and generating speech, music, sound, and talking head [J ] . Proceedings of the AAAI Conference on Artificial Intelligence , 2024 , 38 ( 21 ): 23802 - 23804 .

Liu H T , Li C Y , Wu Q Y , et al . Visual instruction tuning [C ] // Advances in Neural Information Processing Systems 36 . New Orleans : Neural Information Processing Systems Foundation , 2023 : 34892 - 34916 .

OpenAI . GPT-4o system card [EB/OL ] . 2024 : arXiv : 2410 . 21276 . https://arxiv.org/abs/2410.21276 https://arxiv.org/abs/2410.21276

Gemini Team , Google . Gemini 1 . 5 : Unlocking multimodal understanding across millions of tokens of context[EB/OL ] . 2024: arXiv : 2403 . 05530 . https://arxiv.org/abs/2403.05530 https://arxiv.org/abs/2403.05530

Wu S Q , Fei H , Qu L G , et al . NExT-GPT: Any-to-any multimodal LLM [C ] // Proceedings of the 41st International Conference on Machine Learning . Vienna : PMLR , 2024 .

Kojima T , Gu S S , Reid M , et al . Large language models are zero-shot reasoners [C ] // Advances in Neural Information Processing Systems 35 . New Orleans : Neural Information Processing Systems Foundation , 2022 : 22199 - 22213 .

Wang X Z , Wei J , Schuurmans D , et al . Self-consistency improves chain of thought reasoning in language models [EB/OL ] . 2022 : arXiv : 2203 . 11171 . https://arxiv.org/abs/2203.11171 https://arxiv.org/abs/2203.11171

Zhou A , Yan K , Shlapentokh-Rothman M , et al . Language agent tree search unifies reasoning acting and planning in language models [EB/OL ] . 2023 : arXiv : 2310 . 04406 . https://arxiv.org/abs/2310.04406 https://arxiv.org/abs/2310.04406

Yao S Y , Zhao J , Yu D , et al . ReAct: Synergizing reasoning and acting in language models [EB/OL ] . 2022 : arXiv : 2210 . 03629 . https://arxiv.org/abs/2210.03629 https://arxiv.org/abs/2210.03629

OpenAI . OpenAI o1 system card [EB/OL ] . 2024 : arXiv : 2412 . 16720 . https://arxiv.org/abs/2412.16720 https://arxiv.org/abs/2412.16720

Sang J T , Xiao J L , Han J R , et al . Beyond pipelines: A survey of the paradigm shift toward model-native agentic AI [EB/OL ] . 2025 : arXiv : 2510 . 16720 . https://arxiv.org/abs/2510.16720 https://arxiv.org/abs/2510.16720

Schick T , Dwivedi-Yu J , Dessì R , et al . Toolformer: Language models can teach themselves to use tools [C ] // Advances in Neural Information Processing Systems 36 . New Orleans : Neural Information Processing Systems Foundation , 2023 : 68539 - 68551 .

Kang Y , Kim J . ChatMOF: an artificial intelligence system for predicting and generating metal-organic frameworks using large language models [J ] . Nature Communications , 2024 , 15 ( 1 ): 4705 .

Gao Y F , Xiong Y , Gao X Y , et al . Retrieval-augmented generation for large language models: A survey [EB/OL ] . 2023 : arXiv : 2312 . 10997 . https://arxiv.org/abs/2312.10997 https://arxiv.org/abs/2312.10997

Zhao A , Huang D , Xu Q , et al . ExpeL: LLM agents are experiential learners [J ] . Proceedings of the AAAI Conference on Artificial Intelligence , 2024 , 38 ( 17 ): 19632 - 19642 .

Cai T F , Zhang N G , Li Y S , et al . Occam’s razor in AI-assisted complex diagnosis: a comparative effectiveness study of single large language models versus multi-agent systems in resource-constrained primary care settings [R/OL ] . 2026 . https://ssrn.com/abstract=6148075 https://ssrn.com/abstract=6148075

Maia E H B , Medaglia L R , Da Silva A M , et al . Molecular architect: a user-friendly workflow for virtual screening [J ] . ACS Omega , 2020 , 5 ( 12 ): 6628 - 6640 .

Gjerding M , Skovhus T , Rasmussen A , et al . Atomic simulation recipes: a python framework and library for automated workflows [J ] . Computational Materials Science , 2021 , 199 : 110731 .

Yekeen A A , Durojaye O A , Idris M O , et al . CHAPERONg: a tool for automated GROMACS-based molecular dynamics simulations and trajectory analyses [J ] . Computational and Structural Biotechnology Journal , 2023 , 21 : 4849 - 4858 .

Ries B , Alibay I , Anand N M , et al . Automated absolute binding free energy calculation workflow for drug discovery [J ] . Journal of Chemical Information and Modeling , 2024 , 64 ( 14 ): 5357 - 5364 .

Schmitz G , Yönder Ö , Schnieder B , et al . An automatized workflow from molecular dynamic simulation to quantum chemical methods to identify elementary reactions and compute reaction constants [J ] . Journal of Computational Chemistry , 2021 , 42 ( 32 ): 2264 - 2282 .

Singh H , Raja A , Prakash A , et al . Gmx_qk: an automated protein/protein–ligand complex simulation workflow bridged to MM/PBSA, based on gromacs and zenity-dependent GUI for beginners in MD simulation study [J ] . Journal of Chemical Information and Modeling , 2023 , 63 ( 9 ): 2603 - 2608 .

Löhr T , Assante M , Dodds M , et al . Navigating the Maize: cyclic and conditional computational graphs for molecular simulation [J ] . Digital Discovery , 2024 , 3 ( 12 ): 2551 - 2559 .

Ieong P U , Sørensen J , Vemu P L , et al . Progress towards automated Kepler scientific workflows for computer-aided drug discovery and molecular simulations [J ] . Procedia Computer Science , 2014 , 29 : 1745 - 1755 .

Doerr S , Harvey M J , Noé F , et al . HTMD: high-throughput molecular dynamics for molecular discovery [J ] . Journal of Chemical Theory and Computation , 2016 , 12 ( 4 ): 1845 - 1852 .

Hayashi Y , Shiomi J , Morikawa J , et al . RadonPy: automated physical property calculation using all-atom classical molecular dynamics simulations for polymer informatics [J ] . npj Computational Materials , 2022 , 8 ( 1 ): 222 .

Gygli G , Pleiss J . Simulation foundry: automated and F.A.I.R. molecular modeling [J ] . Journal of Chemical Information and Modeling , 2020 , 60 ( 4 ): 1922 - 1927 .

Bayarri G , Andrio P , Hospital A , et al . BioExcel Building Blocks Workflows (BioBB-Wfs), an integrated web-based platform for biomolecular simulations [J ] . Nucleic Acids Research , 2022 , 50 ( W1 ): W99 - W107 .

Liu J X , Zhu T N , Ye C Y , et al . VASPilot: MCP-facilitated multi-agent intelligence for autonomous VASP simulations [J ] . Chinese Physics B , 2025 , 34 ( 11 ): 117106 .

Vriza A , Kornu U , Koneru A , et al . Multi-agentic AI framework for end-to-end atomistic simulations [J ] . Digital Discovery , 2026 , 5 ( 1 ): 440 - 452 .

Mendible-Barreto O A , Díaz-Maldonado M , Carmona Esteva F J , et al . DynaMate: leveraging AI-agents for customized research workflows [J ] . Molecular Systems Design & Engineering , 2025 , 10 ( 7 ): 585 - 598 .

Pham T D , Tanikanti A , Keçeli M . ChemGraph as an agentic framework for computational chemistry workflows [J ] . Communications Chemistry , 2026 , 9 ( 1 ): 33 .

Ghafarollahi A , Buehler M J . AtomAgents: Alloy design and discovery through physics-aware multi-modal multi-agent artificial intelligence [EB/OL ] . 2024 : arXiv : 2407 . 10022 . https://arxiv.org/abs/2407.10022 https://arxiv.org/abs/2407.10022

Zou Y H , Cheng A H , Aldossary A , et al . El Agente: an autonomous agent for quantum chemistry [J ] . Matter , 2025 , 8 ( 7 ): 102263 .

Campbell Q , Cox S , Medina J , et al . MDCrow: Automating molecular dynamics workflows with large language models [J ] . Machine Learning: Science and Technology , 2026 , 7 ( 2 ): 025037 .

Petković M , Menkovski V , Calero S . Towards fully automated molecular simulations: Multi-agent framework for simulation setup and force field extraction [EB/OL ] . 2025 : arXiv : 2509 . 10210 . https://arxiv.org/abs/2509.10210 https://arxiv.org/abs/2509.10210

Ding K Y , Yu J , Huang J J , et al . SciToolAgent: A knowledge-graph-driven scientific agent for multitool integration [J ] . Nature Computational Science , 2025 , 5 ( 10 ): 962 - 972 .

Zhang B H , Li X , Xu H C , et al . TopoMAS: Large language model driven topological materials multiagent system [EB/OL ] . 2025 . https://arxiv.org/abs/2507.04053 https://arxiv.org/abs/2507.04053

Wang Z Q , Huang H S , Zhao H C , et al . DREAMS: Density functional theory based research engine for agentic materials simulation [EB/OL ] . 2025 : arXiv : 2507 . 14267 . https://arxiv.org/abs/2507.14267 https://arxiv.org/abs/2507.14267

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

复合金属CuCr2O4表面CNCl水解反应机理

大模型时代的石油化工行业高质量数据集：挑战与机遇

基于大语言模型的乏燃料后处理脉冲柱萃取过程预测

组分配比影响氯化胆碱-乳酸低共熔溶剂碱木质素溶解度的底层机理研究

基于量子力学分析β-内酰胺合成酶合成吡拉西坦中间体的机制