Biomni: The General-Purpose Biomedical AI Agent Transforming Research

Introduction

In the realm of biomedical research, scientists constantly grapple with challenges like processing massive datasets, designing complex experiments, and accelerating the pace of discovery. Amid these challenges, a groundbreaking solution has emerged: Biomni, a general-purpose biomedical AI agent that promises to redefine how research is conducted. By combining advanced large language model (LLM) reasoning with retrieval-augmented planning and code-based execution, Biomni empowers researchers to enhance productivity and generate testable hypotheses at an unprecedented scale. This comprehensive guide explores every aspect of Biomni—from its core functionality and installation process to community contributions and future developments—providing a complete roadmap for scientists looking to integrate this powerful tool into their workflow.

Understanding Biomni

What is Biomni?

At its core, Biomni is a versatile AI agent designed to autonomously perform a wide range of research tasks across various biomedical subfields. Unlike specialized tools limited to specific functions, Biomni serves as an all-in-one assistant capable of handling everything from experimental design to data analysis. Its power lies in its ability to integrate cutting-edge LLM reasoning with retrieval-augmented planning—meaning it can access and process relevant information to inform its decisions—and code-based execution that translates plans into actionable results.

For researchers, this translates to a tool that doesn’t just assist with specific steps but can guide entire research projects. Whether you’re planning a genetic screen, analyzing single-cell RNA sequencing data, or predicting molecular properties, Biomni can interpret your natural language requests and generate meaningful outputs.

The Advantages of Biomni in Research

Traditional biomedical research often involves tedious, time-consuming processes: sifting through literature, manually designing experiments, and analyzing complex datasets. Biomni streamlines these processes by automating critical steps while maintaining scientific rigor.

• Increased Productivity: By handling time-intensive tasks—like identifying relevant genes or annotating sequencing data—Biomni frees researchers to focus on higher-level scientific questions.
• Hypothesis Generation: The agent’s ability to connect diverse datasets and research findings helps generate novel, testable hypotheses that might otherwise go unnoticed.
• Cross-Disciplinary Capability: Biomni’s general-purpose design allows it to operate across subfields, making it invaluable for interdisciplinary research that bridges genetics, immunology, pharmacology, and more.
• Accessibility: Through natural language processing, Biomni eliminates the need for extensive programming knowledge, putting advanced analytical capabilities in the hands of more researchers.

Getting Started with Biomni

Installation Process

Biomni’s software environment is extensive, but the development team has simplified setup with a single setup.sh script. Follow these steps to get started:

1. Environment Setup: Begin by following the instructions in the biomni_env/README.md file. This document provides detailed guidance on preparing your system for Biomni.

2. Activate the Environment: Once the environment is set up, activate it using:

   conda activate biomni_e1
   

3. Install Biomni: Choose one of two installation methods:

   • Install the latest package via pip:

     pip install biomni --upgrade
     

   • Install directly from the GitHub source for the most recent development version.

4. Configure API Keys: Biomni requires access to certain language model APIs. Configure your keys by adding them to your bash profile (~/.bashrc):

   export ANTHROPIC_API_KEY="YOUR_API_KEY"
   export OPENAI_API_KEY="YOUR_API_KEY"  # Optional if using only Claude
   

This setup ensures Biomni has access to the computational resources and language models needed to perform its tasks effectively.

Basic Usage Examples

Once installed, using Biomni is straightforward. The agent accepts natural language commands, making it easy to integrate into your research workflow. Here’s how to get started in Python:

from biomni.agent import A1

# Initialize the agent with a data path (first run downloads ~11GB data lake automatically)
agent = A1(path='./data', llm='claude-sonnet-4-20250514')

# Execute tasks using plain language
agent.go("Plan a CRISPR screen to identify genes that regulate T cell exhaustion, generate 32 genes that maximize the perturbation effect.")
agent.go("Perform scRNA-seq annotation at [PATH] and generate meaningful hypothesis")
agent.go("Predict ADMET properties for this compound: CC(C)CC1=CC=C(C=C1)C(C)C(=O)O")

These examples demonstrate Biomni’s versatility:

• Experimental Design: The first command shows how Biomni can design a CRISPR screen and identify target genes—tasks that would typically require extensive literature review and data analysis.
• Data Analysis: The second command highlights its ability to process single-cell RNA sequencing data from a specified path and generate hypotheses based on the results.
• Molecular Property Prediction: The third command illustrates Biomni’s capability to analyze chemical structures and predict ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) properties—critical for drug development.

Contributing to Biomni

An Open-Science Initiative

Biomni is more than just a tool—it’s an open-science initiative built on community collaboration. The development team recognizes that advancing biomedical AI requires diverse expertise, which is why they actively encourage contributions from researchers, developers, and domain specialists worldwide.

This collaborative approach ensures Biomni remains adaptable to emerging research needs and integrates the latest advancements from across the biomedical field. Whether you’re a computational biologist, software developer, or laboratory researcher, your input can help shape the future of this platform.

Types of Contributions Welcome

The Biomni community welcomes various forms of contributions to enhance the platform’s capabilities:

• New Tools: Specialized analytical functions and algorithms that address specific research challenges. For example, tools for protein structure prediction or pathway analysis could expand Biomni’s utility.
• Datasets: Curated biomedical datasets and knowledge bases. High-quality, well-annotated data is essential for training and improving Biomni’s performance across different research areas.
• Software Integrations: Connections to existing biomedical software packages. Integrating tools like BLAST, Bowtie, or Cell Ranger would allow Biomni to leverage established resources within its workflow.
• Benchmarks: Evaluation datasets and performance metrics. These help assess Biomni’s accuracy and identify areas for improvement, ensuring the platform meets rigorous scientific standards.
• Educational Resources: Tutorials, use cases, and examples that help other researchers understand how to effectively use Biomni in their work.
• Tool Improvements: Many existing tools within Biomni are still in early stages—optimizations, bug fixes, and replacements for underperforming tools are highly valued.

How to Contribute

Getting involved with Biomni is straightforward, regardless of your technical background:

1. Review the Contributing Guide: The Contributing Guide provides detailed instructions on submitting contributions, including coding standards, documentation requirements, and submission processes.

2. Submit Tools or Datasets: If you have a specific tool, database, or software you’d like to integrate, you can submit your suggestion through this form. The Biomni team will review your proposal and work to implement it.

3. Join the Community: Engage with other contributors through Slack, GitHub discussions, or social media channels (linked at the top of the Biomni documentation). Collaborating with the community can help refine your ideas and ensure your contributions align with the platform’s goals.

Building Biomni-E2: The Next Generation

The Limitations of Biomni-E1

While Biomni-E1 represents a significant step forward, it only scratches the surface of what’s possible in biomedical AI. The current version has constraints in handling highly specialized tasks, integrating with certain laboratory equipment, and processing extremely large or niche datasets. These limitations highlight the need for a more robust, flexible platform that can adapt to the diverse needs of biomedical research.

The Vision for Biomni-E2

To address these limitations, the Biomni team is developing Biomni-E2—a next-generation environment designed with and for the community. The core vision is to create a shared library of standard biomedical actions that researchers worldwide can contribute to and utilize.

This collaborative library would standardize common research tasks, making it easier to replicate experiments, compare results, and build on others’ work. For example, a standardized approach to RNA sequencing analysis or CRISPR design within Biomni-E2 could reduce variability between studies and accelerate scientific consensus.

Benefits of Contributing to Biomni-E2

Contributors to Biomni-E2 have the opportunity to shape the future of biomedical AI while gaining recognition for their work:

• Authorship Opportunities: Contributors who make significant impacts—such as developing 10 or more integrated tools—will be invited as co-authors on upcoming publications in top-tier journals or conferences.
• Recognition: All contributors will be acknowledged in Biomni’s publications and documentation, highlighting their role in advancing the platform.
• Professional Development: Working on Biomni-E2 provides valuable experience in AI development, biomedical informatics, and collaborative research—skills highly sought after in academia and industry.
• Research Advancement: By contributing, you’re helping create a tool that can accelerate discoveries in your own field of study, potentially leading to breakthroughs in areas you care about.

Tutorials and Examples

Biomni 101: Getting Started

For new users, the Biomni 101 tutorial provides a comprehensive introduction to the platform. This interactive notebook covers:

• Basic concepts behind Biomni’s functionality
• Step-by-step guides for common tasks
• Troubleshooting tips for installation and execution
• Examples of how to interpret Biomni’s outputs and integrate them into your research workflow

Whether you’re new to AI or an experienced developer, this tutorial helps you become proficient with Biomni quickly.

Upcoming Tutorials

The Biomni team is developing additional resources to expand users’ capabilities, including:

• Advanced usage guides for complex tasks like multi-omics data integration and longitudinal study analysis
• Specialized tutorials for specific subfields, such as neurobiology, cancer research, and infectious diseases
• Guides for contributing to Biomni’s codebase, even for those with limited programming experience
• Case studies highlighting how researchers have used Biomni to accelerate their work, from hypothesis to publication

These resources will be released as part of Biomni’s ongoing commitment to making advanced AI accessible to all biomedical researchers.

Web Interface: No-Code Access to Biomni

Accessing Biomni Without Programming

Not all researchers are comfortable coding—and Biomni’s developers have ensured the platform remains accessible to everyone through its web interface. Available at biomni.stanford.edu, this no-code platform allows users to interact with Biomni using natural language, just like the Python interface.

The web interface is designed for simplicity: users input their research question or task, upload relevant data if needed, and receive outputs in an easy-to-interpret format. This makes Biomni accessible to laboratory researchers, clinicians, and students who may not have extensive computational training.

Video Demonstration

To help users familiarize themselves with the web interface, the Biomni team has created a video tutorial. Available at this link, the video walks through:

• Navigating the web interface
• Submitting different types of research tasks
• Interpreting results and exporting data
• Troubleshooting common issues

The video provides a visual guide to complement the written documentation, making it easier for new users to get started.

Release Schedule

The Biomni team has outlined a clear roadmap for platform updates, ensuring users know what to expect and can plan their research accordingly:

• Upcoming Releases:

  • 8 real-world research task benchmarks and leaderboards to help users evaluate Biomni’s performance against specific research challenges
  • A detailed tutorial on contributing to Biomni, including step-by-step guides for first-time contributors
  • Baseline agent tutorials that explain the underlying architecture and how to customize Biomni for specific research needs

• Completed Releases:

  • Biomni A1+E1: The initial release, featuring core functionalities for experimental design, data analysis, and molecular property prediction

This transparent schedule helps the community stay informed and engage with development efforts.

Important Notes for Users

Version Differences

It’s important to note that the current release (frozen on April 15, 2025) may differ from the latest version available on the web platform. These differences could include new features, bug fixes, or updated datasets in the web version. Users should check the release notes before starting projects to ensure they’re using the most appropriate version for their needs.

Licensing Information

Biomni itself is released under the Apache 2.0 license, which allows for both academic and commercial use. However, some integrated tools, databases, and software within Biomni may have more restrictive commercial licenses.

Before using Biomni for commercial purposes—such as drug development or diagnostic tool creation—users should:

1. Review the license terms for each integrated component
2. Consult with legal experts to ensure compliance with intellectual property laws
3. Check for any restrictions on commercialization, redistribution, or patenting of results generated using specific tools

This due diligence ensures users avoid legal complications and respect the intellectual property rights of tool developers.

Citing Biomni

If you use Biomni in your research, please cite it using the following format to acknowledge the developers and contributors:

@article{huang2025biomni,
  title={Biomni: A General-Purpose Biomedical AI Agent},
  author={Huang, Kexin and Zhang, Serena and Wang, Hanchen and Qu, Yuanhao and Lu, Yingzhou and Roohani, Yusuf and Li, Ryan and Qiu, Lin and Zhang, Junze and Di, Yin and others},
  journal={bioRxiv},
  pages={2025--05},
  year={2025},
  publisher={Cold Spring Harbor Laboratory}
}

Proper citation helps support the continued development of Biomni and ensures contributors receive appropriate credit for their work.

Conclusion

Biomni represents a paradigm shift in biomedical research, leveraging AI to automate complex tasks, accelerate discovery, and democratize access to advanced analytical tools. By integrating LLM reasoning with retrieval-augmented planning and code-based execution, it provides researchers with a versatile assistant capable of handling diverse tasks across biomedical subfields.

Whether you’re planning a CRISPR screen, analyzing single-cell data, or predicting molecular properties, Biomni streamlines workflows and generates actionable insights. Its open-science model ensures continuous improvement through community contributions, while the upcoming Biomni-E2 promises to expand its capabilities even further.

For researchers looking to enhance productivity, generate novel hypotheses, and stay at the forefront of biomedical innovation, Biomni offers a powerful solution. By joining the Biomni community—whether as a user, contributor, or collaborator—you can play a role in advancing biomedical research and accelerating the path from discovery to application.

As with any new tool, getting started with Biomni may require some initial setup and learning, but the investment quickly pays off in time saved and new research opportunities. Explore the tutorials, join the community, and experience firsthand how AI can transform your research workflow.

The future of biomedical research is collaborative, data-driven, and AI-enhanced—and Biomni is leading the way.