pharmolix/openbiomed

English | 中文

Feel free to use our Agent Platform for Biomedicine and Life Science at this website!

News 🎉

• [2026/05/01] 🔥 CodeFP (📃Paper) has been released.

CodeFP is a co-generative PLM framework jointly released by PharMolix Inc. and the Institute of AI Industry Research (AIR), Tsinghua University. It unifies sequence and structure generation to advance de novo functional protein design. Empirical evaluations demonstrate that CodeFP achieves state-of-the-art performance in both functional consistency and structural foldability on de novo functional protein design benchmarks.

Feel free to try out the functional-protein-design skill to experience CodeFP.

• [2026/03/20] 🔥 We release OpenBioMed Skills, a comprehensive collection of 45 skills for biomedical research and drug discovery empowered by Claude Code.

OpenBioMed Skills is a comprehensive skill set released jointly by PharMolix and Institute of AI Industry Research (AIR), Tsinghua University. It provides users with end-to-end solutions for complicated biomedical research tasks spanning drug discovery, protein analysis & engineering, and single-cell omics data analysis. It also presents a copilot mode for creating your own skills by interacting with an LLM agent and the OpenBioMed toolkits. Feel free to have a quick try and investigate our skills.

Table of contents

• Introduction
• Installation
• Quick Start with Claude Code
• Tutorials
• Previous version
• Limitations
• Cite us

Introduction

This repository holds OpenBioMed, a comprehensive skillset and deep learning toolkit for biomedical discovery. The goal of OpenBioMed is to help researchers build and use AI-driven workflows for solving complicated biomedical research tasks. Enpowered by Claude Code, OpenBioMed provides 45 skills that provides end-to-end solutions for complicated biomedical research tasks. OpenBioMed builds 20+ tools that covers a wide range of downstream applications, facilitating the construction of your own skills with a seamless user-agent interactions.

OpenBioMed provide researchers with access to:

• 45 skills that provides end-to-end solutions for complicated biomedical research tasks, spanning drug discovery, protein analysis & engineering, single-cell omics data analysis, and data retrieval & knowledge.
• 4 types of data modalities: OpenBioMed provide easy-to-use APIs for researchers to access and process different types of data including molecules, proteins, pockets, and texts.
• 20+ tools powered by deep learning models, comprising exclusive models such as PharmolixFM, BioMedGPT-R1, BioMedGPT and MutaPLM.

Here is a list of currently available skills. This is a continuing effort and we are working on further growing the skillset.

If you are interested in the tools that OpenBioMed skills are built on, please check out the following list.

Installation

To enable basic features of OpenBioMed, please execute the following:

conda create -n OpenBioMed python=3.9
conda activate OpenBioMed
pip install torch==1.13.1+{your_cuda_version} torchvision==0.14.1+{your_cuda_version} torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/{your_cuda_version}  
pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-1.13.1+{your_cuda_version}.html
pip install pytorch_lightning==2.0.8 peft==0.9.0 accelerate==1.3.0 --no-deps -i https://pypi.tuna.tsinghua.edu.cn/simple
pip install -r requirements.txt

We recommend using cuda=11.7 to set up the environment. Other versions of cudatoolkits may lead to unexpected problems.

To enable visualization tools and vina score computation tools, you should install the following packages:

# For visualization
conda install -c conda-forge pymol-open-source
pip install imageio

# For AutoDockVina
pip install meeko==0.1.dev3 pdb2pqr vina==1.2.2 
python -m pip install git+https://github.com/Valdes-Tresanco-MS/AutoDockTools_py3

# For PoseBusters
pip install posebusters==0.3.1

# For overlap-based evaluation
pip install spacy rouge_score nltk
python
>>> import nltk
>>> nltk.download('wordnet')
>>> nltk.download('omw-1.4')

# For LangCell
pip install geneformer

After downloading the dependencies, you can run the following command to install the package and use our APIs more conveniently:

pip install -e .
# Try using OpenBioMed APIs
python
>>> from open_biomed.data import Molecule
>>> molecule = Molecule(smiles="CC(=O)OC1=CC=CC=C1C(=O)O")
>>> print(molecule.calc_logp())

Build Docker

Executing ./scripts/docker_run.sh directly will build the Docker image and run the container, launching the backend services on ports 8082 and 8083.

sh ./scripts/docker_run.sh

At the same time, we also provide a pre-built docker image, which can be pulled and used directly.

Quick Start with Claude Code

OpenBioMed Skills requires Claude Code to be installed and running.

mkdir .claude
# Install to your workspace skills directory
cp -r skills/* /skills/
claude

• Type /target-based-lead-design: Configure the target protein or disease (e.g. EGFR) and the desired properties of the lead molecule and receive a bunch of diverse lead candidates with a comprehensive report and visualization after a coffee break!
• Type /functional-protein-design: Give your desired functions (e.g. bacteria degradation), let the model generate a functional protein sequence and its 3D structure.
• Type /biomed-skill-creator: Condense and streamline your workflow into a skill by chatting with an LLM agent.

Tutorials

Checkout our Jupytor notebooks for more tutorials!

Other Versions

If you hope to use the features of the previous version, please switch to the v1.0 branch of this repository by running the following command:

git checkout v1.0

We have also provided a nightly version of OpenBioMed with MCP support. You can try it by running the following command:

git checkout mcp

Limitations

This repository holds BioMedGPT-LM-7B, BioMedGPT-10B, and BioMedGPT-R1, and we emphasize the responsible and ethical use of these models. BioMedGPT should NOT be used to provide services to the general public. Generating any content that violates applicable laws and regulations, such as inciting subversion of state power, endangering national security and interests, propagating terrorism, extremism, ethnic hatred and discrimination, violence, pornography, or false and harmful information, etc. is strictly prohibited. BioMedGPT is not liable for any consequences arising from any content, data, or information provided or published by users.

License

This repository is licensed under the MIT License. The use of BioMedGPT-LM-7B and BioMedGPT-10B models is accompanied with Acceptable Use Policy.

Contact Us

We are looking forward to user feedback to help us improve our framework. If you have any technical questions or suggestions, please feel free to open an issue. For commercial support or collaboration, please contact [email protected].

Cite Us

If you find our open-sourced code and models helpful to your research, please consider giving this repository a 🌟star and 📎citing our research papers. Thank you for your support!

To cite OpenBioMed:

@misc{OpenBioMed_code,
      author={Luo, Yizhen and Yang, Kai and Fan, Siqi and Hong, Massimo and Zhao, Suyuan and Chen, Xinrui and Nie, Zikun and Luo, Wen and Xie, Ailin and Liu, Xing Yi and Zhang, Jiahuan and Wu, Yushuai and Nie, Zaiqing},
      title={Code of OpenBioMed},
      year={2023},
      howpublished={\url{https://github.com/Pharmolix/OpenBioMed.git}}
}

To cite BioMedGPT:

@article{luo2024biomedgpt,
  title={Biomedgpt: An open multimodal large language model for biomedicine},
  author={Luo, Yizhen and Zhang, Jiahuan and Fan, Siqi and Yang, Kai and Hong, Massimo and Wu, Yushuai and Qiao, Mu and Nie, Zaiqing},
  journal={IEEE Journal of Biomedical and Health Informatics},
  year={2024},
  publisher={IEEE}
}

To cite PharmolixFM:

@article{luo2025pharmolixfm,
  title={PharMolixFM: All-Atom Foundation Models for Molecular Modeling and Generation},
  author={Luo, Yizhen and Wang, Jiashuo and Fan, Siqi and Nie, Zaiqing},
  journal={arXiv preprint arXiv:2503.21788},
  year={2025}
}

To cite MolFM:

@misc{luo2023molfm,
      title={MolFM: A Multimodal Molecular Foundation Model}, 
      author={Yizhen Luo and Kai Yang and Massimo Hong and Xing Yi Liu and Zaiqing Nie},
      year={2023},
      eprint={2307.09484},
      archivePrefix={arXiv},
      primaryClass={q-bio.BM}
}

To cite LangCell:

@misc{zhao2024langcell,
      title={LangCell: Language-Cell Pre-training for Cell Identity Understanding}, 
      author={Suyuan Zhao and Jiahuan Zhang and Yizhen Luo and Yushuai Wu and Zaiqing Nie},
      year={2024},
      eprint={2405.06708},
      archivePrefix={arXiv},
      primaryClass={q-bio.GN}
}

To cite MutaPLM:

@article{luo2025mutaplm,
  title={MutaPLM: Protein Language Modeling for Mutation Explanation and Engineering},
  author={Luo, Yizhen and Nie, Zikun and Hong, Massimo and Zhao, Suyuan and Zhou, Hao and Nie, Zaiqing},
  journal={Advances in Neural Information Processing Systems},
  volume={37},
  pages={79783--79818},
  year={2025}
}