PharMolix

We are PharMolix, a startup incubated by the Institute for AI Industry Research (AIR), Tsinghua University. We focus on accelerating drug discovery

[PharMolix @PharMolix]

OpenBioMed Skills: Open-Source AI Tools for Biomedical Research 🦀Open-source library encoding biomedical expert knowledge into executable code. Enables AI agents to run 45+ computational tasks via simple chat (no coding needed). GitHub: https://t.co/FGsgobakTh

[PharMolix @PharMolix]

Here is a demo of OpenBioMed Skills running in #Claude Code. This example shows how to quickly build a professional #skill for discovering a diverse set of lead compounds for a specific disease.

[PharMolix @PharMolix]

The "OpenClaw" for Biomedicine is here!

We’re thrilled to announce OpenBioMed Skills — a joint work by Tsinghua DAIR Lab and @PharMolix.

It is the world’s first open-source library that encodes expert biomedical decision-making into executable code.

Why it matters:
Traditional drug R&D requires heavy engineering. OpenBioMed Skills changes the game. It empowers researchers to build and run end-to-end workflows through a simple chat interface — zero advanced coding required.

What’s under the hood?
Built on the #OpenBioMed agent platform, it gives your AI agents "hands" to execute 45+ professional skills:

Drug Discovery: ADMET, Retrosynthesis, Lead Design
Protein Engineering: Functional Protein Design, Mutation Analysis.
Single-Cell Omics: scRNA-seq Analysis and Single-cell Foundation Models.
Knowledge Access: Seamless integration with PubChem, UniProt, ChEMBL, and more databases.

Forget manual pipeline construction. With OpenBioMed Skills, a simple chat interface is all you need to trigger autonomous AI research workflows.

Join the movement. Let’s build the future of AI for Science together!
💻 GitHub: [https://t.co/yDp0ExmOGB]

#AI4Science #BioTech #OpenSource #DrugDiscovery #LLM #OpenBioMed #Tsinghua #OpenClaw #Claude

[PharMolix @PharMolix]

The expert version for retrosynthetic planning will be released on OpenBioMed upon publication. Please stay tuned for updates.

[PharMolix @PharMolix]

A strong step toward molecular R&D agents: Tsinghua AIR and @PharMolix have open-sourced BioMedGPT-Mol, a molecular foundation model. It achieves SOTA across 6 task categories, spanning both molecular understanding and generation. Building on this foundation, a further fine-tuned version reaches SOTA performance on RetroBench as an end-to-end retrosynthetic planner. It has the potential to serve as a backbone for next-generation molecular R&D agents.
The paper is now available at https://t.co/3VZgi2neUe
The foundation model is open-sourced at https://t.co/yDp0Exnmw9

@ZaiqingNie

[PharMolix @PharMolix]

A major step toward the AI Virtual Cell: Tsinghua AIR & @PharMolix open-sourced SToFM, a new multi-scale foundation model for spatial transcriptomics. Unlike models focused on only one level of biology, SToFM learns across:
🧬 gene expression
🧫 cell neighborhoods
🧠 tissue architecture。 Pretrained on 88M cells, it delivers strong results on key spatial biology tasks and helps move AI from single-cell understanding to true multicellular, spatially aware modeling.
* The paper has been accepted for publication at ICML2025（https://t.co/ZXvXTJRVqQ）
* The code is now available at https://t.co/fHGzXEmEA7. #ICML2025 # SpatialTranscriptomics #VirtualCell #AIBiology