“Curing All Disease by next century is too conservative" - Mark Zuckerberg
Summary
Biohub, a philanthropic initiative by Mark Zuckerberg and Priscilla Chan, is dedicated to accelerating scientific progress to cure, prevent, and manage all diseases, now aiming for a timeline considered "too conservative" at the century's end. Committing \$500 million to its virtual biology initiative, Biohub integrates frontier AI with frontier biology, generating unique datasets to model biological systems hierarchically from proteins to cells. The organization emphasizes open-source tools, exemplified by its role in funding single-cell sequencing for the Human Cell Atlas and developing CellByGene. A recent major launch is ESM Fold, an open-source protein language model trained on billions of protein sequences. ESM Fold achieves state-of-the-art atomic-resolution protein structure prediction, including protein-protein and protein-antibody interactions, and has been used to fold over 1.1 billion proteins and digitally design therapeutic-grade single-chain antibodies. Biohub's future vision includes developing a "virtual cell" model, enhancing mechanistic interpretability, and enabling personalized medicine by predicting off-target drug effects.
Key takeaway
For AI Scientists and Directors of AI/ML aiming to accelerate therapeutic discovery, Biohub's model highlights a critical shift: integrating frontier AI with novel biological data generation, then distributing tools open-source. This approach, exemplified by ESM Fold's ability to digitally design therapeutic-grade antibodies, dramatically shortens design cycles and enhances predictability for personalized medicine. You should explore adopting open-source biological AI models and consider investing in internal "frontier biology" capabilities to generate unique datasets, thereby accelerating your therapeutic pipelines and enabling more precise, individualized interventions.
Key insights
Integrating frontier AI with frontier biology and open-source tools accelerates disease understanding and personalized therapeutic design.
Principles
- Open-source tools maximize scientific impact and community engagement.
- Hierarchical biological modeling from proteins to cells is essential.
- Mechanistic interpretability extracts new biological knowledge from AI models.
Method
Generate novel biological data through advanced wet lab techniques, train large-scale AI models on this data, and release tools open-source to the scientific community.
In practice
- Apply protein language models for rapid protein structure prediction and design.
- Validate AI-designed molecules with targeted wet lab experiments.
- Utilize single-cell atlases to predict potential drug off-target effects.
Topics
- Biohub
- Frontier AI
- Frontier Biology
- Protein Language Models
- Open Science
- Personalized Medicine
Best for: AI Scientist, Research Scientist, Director of AI/ML
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by No Priors: AI, Machine Learning, Tech, & Startups.