A world model for proteins is here

2026-05-28 · Source: The Rundown AI · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Life Sciences & Biology, Emerging Technologies & Innovation · Depth: Fundamental Awareness, medium

Summary

Biohub, backed by Mark Zuckerberg and Priscilla Chan's CZI, has released a "world model of protein biology" featuring new Evolutionary Scale Models (ESM). The core component, ESMFold2, is a protein language model (ESMC) trained on 2.8 billion sequences, designed to predict protein structure and design new proteins. ESMFold2 claims state-of-the-art performance in structure prediction, including protein-protein and antibody-antigen interactions, outperforming AlphaFold. Early lab results show it designing binders against five cancer and immune targets with hit rates between 36% and 88%. The initiative also includes ESM Atlas, a map of 6.8 billion protein sequences and 1.1 billion predicted structures, revealing novel evolutionary connections. This open-source stack, supported by a \$500M Virtual Biology Initiative, aims to accelerate drug discovery. Additionally, OpenAI Foundation committed \$250M to address AI-driven economic disruption, and Trajectory launched with \$15M to develop AI that continuously learns from user feedback.

Key takeaway

For research scientists and biotech firms focused on drug discovery, Biohub's open-source protein "world model" offers a powerful new foundation. You should explore ESMFold2 and ESM Atlas to accelerate protein mapping, prediction, and design, potentially reducing drug discovery timelines from years to months. Additionally, consider how continuous learning AI, like Trajectory's platform, could enhance your internal AI tools by improving them with real-world feedback.

Key insights

Biohub's open-source protein "world model" (ESMFold2, ESM Atlas) accelerates drug discovery by mapping, predicting, and designing proteins.

Principles

• Open-source models democratize drug discovery.
• Continuous learning improves AI over time.
• AI can mitigate economic disruption.

Method

To teach AI agents editing style: create project folders, prompt for editorial rules, draft/snapshot, edit, compare, and update rules.

In practice

• Use ESMFold2 for protein structure prediction.
• Implement continuous post-training for AI models.
• Automate rule updates from edited content.

Topics

• Protein Engineering
• Drug Discovery
• AI Models
• Continual Learning
• Economic Impact of AI
• Open-Source AI

Best for: AI Scientist, General Interest, Research Scientist, Director of AI/ML

Related on AIssential

• Building the blocks of life — Integrating physics and geometric deep learning enhances protein modeling, improving generalizability and reducing data needs.
• Zuckerberg's $500M AI biology swing — Large-scale data generation and advanced AI models are converging to revolutionize disease understanding and food science.
• Why We're Not Close to AI Drugs Yet | Isomorphic Labs — Drug design comprises multiple distinct ML problems, not a single one, despite AlphaFold 2's protein structure prediction success.
• Bringing AI-driven protein-design tools to biologists everywhere — Democratizing AI-driven protein design tools empowers biologists to accelerate drug discovery and engineer novel proteins.
• 🔬ESM: The Bitter Lesson is Coming for Proteins - Alex Rives, BioHub — Scaling protein language models with vast, diverse data reveals emergent biological capabilities and enables programmable design.
• Meta Pivots From Open Weights, Big Pharma Bets On AI, Regulatory Patchwork, Simulating Human Cohorts — AI-native teams, advanced models, and AI-driven drug discovery are reshaping industries, while regulatory landscapes become increasingly complex.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Rundown AI.