I promised more on TMol in the RFD4-Proteina thread. Now with the paper on arxiv here we go! 🧵
Rosetta-style all-atom energies, GPU batching, and gradients in PyTorch. A way to put molecular energetics directly into your model, combining deep learning with physics. 1/n
1/18 🧵A thread about the evolution of four generative AI models for biomolecular design (Proteína, La-Proteína, Proteína-Complexa and RFD4-Proteina), which makes me think we're moving toward fully programmable protein design 🔥
I’ll be in my hometown Guntur sometime next week for few days. Anyone on my TL from there and are into AI research? I’d love to chat and offer any guidance. 🙂
(Appreciate retweets to reach some right folks!)
Excited to share what we’ve been building with Chaperone!
In our benchmark, Chaperone reduced token usage by 86%. We also went from hypothesis to sequences ready for wet-lab testing in just two weeks—yielding a 1.32 nM cortisol binder in the first experimental round!
This is what a communal brain—a central tenet of the IPD approach—looks like in action. 🧠
We're thrilled to present RFD4-Proteina, co-developed with @NVIDIAHealth! Can't wait to see what you build: https://t.co/ezw5LPM0Bf
Protein design doesn’t have to stop at the canonical 20 amino acids! After our work on phosphotyrosine binders, RFD4-Proteina adds phosphoserine and phosphothreonine (below) to the repertoire, with binders that distinguish adjacent H3pS10 and H3pT11 sites in the histone H3 tail!
🧬 Design proteins faster and send fewer, better candidates to the lab.
Introducing RFD4-Proteina, a new NVIDIA BioNeMo open model co-developed with @UWproteindesign.
▫️Fast: 13× faster generation than RFdiffusion3 at 200 residues
▫️ Smart: state-of-the-art across 12+ protein design tasks
▫️ Programmable: set your design inputs and use test-time compute on harder problems
Start with our binder-design tutorial https://t.co/fmS7miF1VD
Great things come from great teams! Super excited to share https://t.co/TnVV8YAW3f, a step-function improvement in generative protein design! A year in the making with @UWproteindesign , NVIDIA GenAIR and the amazing @DidiKieran@nscorley and @RafiBrent! 🧵1/n
What a wild ride! After a fantastic collaboration between @UWproteindesign and Nvidia GenAIR, we've released the latest in the RFdiffusion and Proteina lineages, namely... RFD4-Proteina! Computationally and in the lab, the results have astonished us🧵 1/n https://t.co/PPddK8nlsb
1/n
Crossover episode! We introduce RFD4-Proteina, our latest design model - and the result of a year-long collaboration between the Baker Lab and NVIDIA GenAIR (the name says it all…). I hope you’ll agree - the results are astonishing. 🌐 Website: https://t.co/FRbZU8OiKy
Avengers assemble: 1 year BakerLab + NVIDIA GenAIR = 🔥 https://t.co/Cd2qiugCJo
Presenting RFD4-Proteina, our latest protein design model, significantly better and faster than everything we compared against, both in silico and in the lab! 1/30
Today, RFdiffusion3 is out in @ScienceMagazine!
https://t.co/hITqU9p4Ot
We have many exciting updates to share! We're also releasing Chaperone, an agent harness for general bio-molecular design. 1/n
✨Introducing Prompteina: all-atom protein design directly from natural language✨
- Control diverse protein properties through a single text prompt
- Higher predicted binding confidence than baselines
- First experimentally validated de novo binders reported for GM2A
🧵
With the help of @huggingface we (/w @RisingSayak) are building a ML Club India 🇮🇳
What we want to do:
1. Online talks
2. IST compatible timing
2. Open to all
More to come in this week! Watch this space. 🤗
Special thanks to @LysandreJik who motivated me to keep working on this. 🔥
@gowthami_s It was lovely chatting with you regarding AI for Materials Discovery! Thanks for the cool questions and explaining your work on image generation.
Some points that stood out from hallway discussion with @JeffDean on continual learning:
1. RL/Finetuning with user preferences at deployment could be dangerous as we cannot control the agent's behavior.
2. We need new paradigms forwards for pretraining to enable the agentic continual leanring.
3. Google is not seeing the slowdown of the current pretraining scaling law though.
All the points above are not mutually exclusive. We need more on-the-ground technical discussions than glossing over the headlines.