Mingan Chen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Mingan Chen has a total of 23 co-authors. Here are the top co-authors by number of shared publications:
- Duanhua Cao
University of Chinese Academy of Sciences
2 shared publications - Jiacheng Xiong
Chinese Academy of Sciences
2 shared publications - Mingyue Zheng
Shanghai Institute of Materia Medica, Chinese Academy of Sciences
2 shared publications - Xutong Li
Drug Discovery and Design Center, State Key Laboratory of Drug Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai, China
2 shared publications - Buying Niu
Drug Discovery and Design Canter, State Key Laboratory of Drug Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, 555 Zuchongzhi Road, Shanghai 201203, China
1 shared publication - Dan Teng
Drug Discovery and Design Center, State Key Laboratory of Drug Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai, China
1 shared publication
Latest Publications
- Open Access
- 22 October 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Deep Residual Learning for Molecular Force Fields
Accurate interatomic interaction modeling via force fields is critical for biological studies and drug discovery, but balancing accuracy and generalizability remains challenging. We introduce ResFF (Residual Learning Force Field), a hybrid machine learning force field that employs deep residual learning to integrate physics-based learnable molecular mechanics covalent terms with residual corrections from a lightweight equivariant neural network. Through a three-stage joint optimization, the two components are trained in a complementary manner to achieve optimal performance. Benchmarks show ResFF outperforms classical and neural network force fields in generalization (mean absolute error (MAE): 1.16 kcal/mol on Gen2-Opt, 0.90 kcal/mol on DES370K), torsional profiles (MAE: 0.45/0.48 kcal/mol on TorsionNet-500 and Torison Scan), and intermolecular interactions (MAE: 0.32 kcal/mol on S66×8). It also enables precise energy minima reproduction and stable molecular dynamics of biological systems. ResFF merges physical constraints with neural expressiveness, offering a robust tool for accurate and efficient molecular simulation.
- 22 October 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessDeep Residual Learning for Molecular Force Fields
Accurate interatomic interaction modeling via force fields is critical for biological studies and drug discovery, but balancing accuracy and generalizability remains challenging. We introduce ResFF (Residual Learning Force Field), a hybrid machine learning force field that employs deep residual learning to integrate physics-based learnable molecular mechanics covalent terms with residual corrections from a lightweight equivariant neural network. Through a three-stage joint optimization, the two components are trained in a complementary manner to achieve optimal performance. Benchmarks show ResFF outperforms classical and neural network force fields in generalization (mean absolute error (MAE): 1.16 kcal/mol on Gen2-Opt, 0.90 kcal/mol on DES370K), torsional profiles (MAE: 0.45/0.48 kcal/mol on TorsionNet-500 and Torison Scan), and intermolecular interactions (MAE: 0.32 kcal/mol on S66×8). It also enables precise energy minima reproduction and stable molecular dynamics of biological systems. ResFF merges physical constraints with neural expressiveness, offering a robust tool for accurate and efficient molecular simulation.
- Open Access
- 31 January 2024
- Published by American Chemical Society (ACS) in ChemRxiv
Fine-tuning Large Language Models for Chemical Text Mining
Extracting knowledge from complex and diverse chemical texts is a pivotal task for both experimental and computational chemists. The task is still considered to be extremely challenging due to the complexity of the chemical language and scientific literature. This study explored the power of fine-tuned large language models (LLMs) on five intricate chemical text mining tasks: compound entity recognition, reaction role labelling, metal-organic framework (MOF) synthesis information extraction, nuclear magnetic resonance spectroscopy (NMR) data extraction, and the conversion of reaction paragraph to action sequence. The fine-tuned LLMs models demonstrated impressive performance, significantly reducing the need for repetitive and extensive prompt engineering experiments. For comparison, we guided GPT-3.5 and GPT-4 with prompt engineering and fine-tuned GPT-3.5 as well as other open-source LLMs such as Llama2, T5, and BART. The results showed that the fine-tuned GPT models excelled in all tasks. It achieved exact accuracy levels ranging from 69% to 95% on these tasks with minimal annotated data. It even outperformed those task-adaptive pre-training and fine-tuning models that were based on a significantly larger amount of in-domain data. Given its versatility, robustness, and low-code capability, leveraging fine-tuned LLMs as flexible and effective toolkits for automated data acquisition could revolutionize chemical knowledge extraction.
- 31 January 2024
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessFine-tuning Large Language Models for Chemical Text Mining
Extracting knowledge from complex and diverse chemical texts is a pivotal task for both experimental and computational chemists. The task is still considered to be extremely challenging due to the complexity of the chemical language and scientific literature. This study explored the power of fine-tuned large language models (LLMs) on five intricate chemical text mining tasks: compound entity recognition, reaction role labelling, metal-organic framework (MOF) synthesis information extraction, nuclear magnetic resonance spectroscopy (NMR) data extraction, and the conversion of reaction paragraph to action sequence. The fine-tuned LLMs models demonstrated impressive performance, significantly reducing the need for repetitive and extensive prompt engineering experiments. For comparison, we guided GPT-3.5 and GPT-4 with prompt engineering and fine-tuned GPT-3.5 as well as other open-source LLMs such as Llama2, T5, and BART. The results showed that the fine-tuned GPT models excelled in all tasks. It achieved exact accuracy levels ranging from 69% to 95% on these tasks with minimal annotated data. It even outperformed those task-adaptive pre-training and fine-tuning models that were based on a significantly larger amount of in-domain data. Given its versatility, robustness, and low-code capability, leveraging fine-tuned LLMs as flexible and effective toolkits for automated data acquisition could revolutionize chemical knowledge extraction.