AI & Tech
BioHub Folds 1.1 Billion Proteins Without Designing Model for Specific Diseases
No Priors Podcast
Biohub: The Future of Biology is Open-Source with Co-Founders Mark Zuckerberg, Priscilla Chan, and Head of Science Alex Rives
"We folded over 1.1 billion proteins and predicted their structures, and we didn't design a model for antibodies. We didn't design a model to be able to bind one particular target. We just designed a model that could understand proteins. You get protein design as an emergent property."
Alex Reeves disclosed that BioHub's ESMFold system achieved state-of-the-art results across all structure prediction benchmarks, especially protein-protein and protein-antibody interactions critical for therapeutics, using a general-purpose model rather than disease-specific training. The protein design capabilities emerged spontaneously from the language model architecture, suggesting fundamental biology can be learned without explicit programming.
From this episode
No Priors Podcast