Your workflows need more than just instructions.
Introducing OpenBio Agents. Your AI co-scientists, each mastering a specific workflow, so you can focus on discovery.
You can even create your own that fit your expertise!
Check it out:
Are you interested/building in AIxBio? Are you in Bangalore?
Me, @anaturalist and @vinayprabhu are organising "AIxBio BLR" at IISc this Saturday, April 25th at 11:30 AM. The goal? Build the community that is missing and help each other in any way possible.
We have Founders, PhDs, Faculty, PostDocs attending the event. If you are building something in the space or have a unique expertise, we would love to have you.
Register on the link below.
UPDATE: LLMs in biology still have some way to go.
A few weeks ago, I had said that @AnthropicAI models are the best at creating publication level figures. But while adding image support to PyMolAI, I realized that is not the case.
Left: A simple pose from a paper I wanted to recreate.
Right: What Sonnet 4.6 was able to make.
They don't look similar at all.
"instead of knowing what a good pose is, the models just know the steps to get to a good pose"
What this means is, that the models don't understand molecular/protein structures. They just have been trained on enough pymol commands to get to a good point. Anthropic models more so than others.
I tested models from @AnthropicAI@OpenAI@Google@Zai_org@MiniMax_AI and @Kimi_Moonshot on whether they can create a publication-level view of a protein-ligand binding site and the results were surprising.
TL;DR: Anthropic models did the best and Gemini models did the worst.
Task was simple:
"Load 5DEL and create a publication ready view of the ORO and FMN binding site"
Here are the results:
1. Sonnet 4.6 from @AnthropicAI:
- Fast
- Used the tools pretty well
- Created the best view with clear labelling and color choices
- always felt in control of what it was going for
- very impressed
I am Open-Sourcing PyMolAI!
Meet PyMolAI, an AI agent that can talk to your protein structures.
Built on top of PyMOL, PyMolAI lets you interact with your structures in plain language. Whether you're:
- Analyzing protein structures
- Aligning complexes
- Creating publication-ready figures
- Or running design workflows
PyMolAI interprets your request, executes the necessary PyMOL commands, and manages the workflow for you.
It integrates with @OpenBioAI APIs, giving you access to tools like Boltz, ProteinMPNN, and BoltzGen — directly from your PyMOL session.
It has local chat history with session syncing, so you can pick up exactly where you left off.
We have updated the default MAX model on @OpenBioAI from Sonnet 4 (which I really liked) to Sonnet 4.5, because of the API issues from Anthropic's side. There were a lot of errors when using Sonnet 4. I feel they're realigning resources for newer models.
Sonnet 4.5 is pretty solid, tends to take more notes. But should feel a lot more smoother, have been testing it a lot past couple of days. Anyways if you tried out OpenBio in the past couple of days and faced issues, please give it a try again. Lots of cool stuff otw and already there.
OpenBio viewers update!
I had this one in mind since I started @OpenBioAI , but when @ATinyGreenCell said that I should add it, I had to.
Now you can view your ✨pretty plasmids✨ in OpenBio. Openbio supports all the common formats like Genbank and SnapGene.
This update comes with improvement to existing fasta file viewers along with a new MSA viewer. Have a look:
Announcing OpenBio skills and APIs for your Agents and Molts🦀!
Over the past few days we've seen how far agents have come, they're doing everything from making payments to making religions.
Now we want to empower them to do biology research. From drug discovery to literature search, get your agents to do anything for you.
Works with any of your favourite agent platform @openclaw@mograxyz@claudeai@opencode and more! Set it up now!
Building OpenBio is so fun! This feature has been on prod since December but we only got around to talking about it now. So much more is already there that we have not talked about and so much more is on the way!