NVIDIA releases open-source BioNeMo Agent Toolkit to build AI agents for protein prediction and molecular docking

Original post

Proxima@proximabio

Discovering proximity-based medicines means connecting many disciplines such as foundation model training, chemistry, biology, and proteomics. Powered by the NVIDIA BioNeMo Agent Toolkit, our AI agents tie these domains together, turning new biology into therapeutic hypotheses faster and moving promising programs toward patients sooner. We're proud to build that future alongside @NVIDIAHealth as they launch the BioNeMo Agent Toolkit at BIO 2026.

Links: - BioNeMo Agent Toolkit GitHub Repo: https://bit.ly/43Qp8ja - Blogpost: https://bit.ly/4vssVPG

7:31 AM · Jun 23, 2026 · 1.4K Views

VIEWS51.3KBOOKMARKS18LIKES133RETWEETS5REPLIES17

Martin Shkreli@MartinShkreli

This x100000

Josie Zayner@josiezayner

No scientist ever said, "you know what I need? An AI agent to do molecular docking and protein structure prediction."

Why are we pretending this is progress?

4h51.3K13318

Cooker.hl | 版本之子 (Theo Arc)@CookerFlips

@MartinShkreli Tbf nobody ever said we needed an AI agent for anything 5 years ago yet here we are

4h8984

Jean-Paul Tres@JPTres

@MartinShkreli I was at thousand genomes and remember they said in front of everyone they want an imputation server. Basically the same thing. I would’ve gladly shared what I was working on, but they don’t really want that. They want some big company to figure out the hard part so they can talk

4h1721

shourya⚠️@flaczip

@MartinShkreli lmao real

4h2211

ali joy@16p11dot2ds

@MartinShkreli idk or care. i finished already.

4h205

samir huseynov@huseynovecon

@MartinShkreli but they will say it if they’re not already

4h69

Facts Merchant@FactsMerchant

@MartinShkreli People excited about it are presumably unaware that docking programs have been around for decades.

3h50

Jean-Paul Tres@JPTres

@MartinShkreli Then whine at lunch about how everything sucks. At my table, it was Mayo Clinic being dismissive of 23 and Me as a public safety hazard after I brought it up and didn’t offer any alternatives. There are lots of problems in research, to do the academic hand wavy non statement too.

4h9

Emily Researcher@potsstrongem

@MartinShkreli No scientist asked for this. We need funding and infrastructure, not another AI toolkit. And let's be real, funding is still woefully inadequate.

4h1