Anthropic has not solved the peptide-binder design problem, but maybe bi[o]hub has

The protein design field is getting pushed forward by human experts in protein design, who would have thought

In my recent post about Claude’s protein design efforts, I said that I was looking for a 2–3 sentence explanation of how Claude consistently did better than the collective field of human protein engineers, when it was using the same tools that everybody else is using. Well, it looks like we have our answer. The answer is ESMFold2 from bi[o]hub. ESMFold2 is a new, open source protein folding model, similar to AlphaFold3, but newer. And faster. And also, apparently, better, in particular for evaluating binders. ESMFold2 was only released a few months ago, so it’s not surprising that it’s not…

saved by

• ochre

related reading

• One-shot design of functional protein binders with BindCraftnature.com
• Protein World Model | Biohubbiohub.ai
• Highly accurate protein structure prediction with AlphaFoldnature.com
• Minibinder design is just not that hardblog.escalante.bio
• AlphaFold2 @ CASP14: “It feels like one’s child has left home.” << Some Thoughts on a Mysterious Universemoalquraishi.wordpress.com
• GitHub - Biohub/esmgithub.com
• Language models of protein sequences at the scale of evolution enable accurate structure predictionbiorxiv.org
• prompts/prompts/multi_target_binder_design_prompt.md · Anthropic/claude-protein-binder-design at mainhuggingface.co
• How AI Revolutionized Protein Science, but Didn’t End It | Quanta Magazinequantamagazine.org
• AlphaProteo generates novel proteins for biology and health researchdeepmind.google
• Engineering Plastic-Degrading Enzymes and PCSK9 Binders with Protein AI Toolsblog.latch.bio
• AlphaFold: a solution to a 50-year-old grand challenge in biologydeepmind.com