Biohub releases a world model of protein biology (biohub.org)

Biohub has launched an innovative open-source platform focused on protein biology, introducing a world model that integrates predictive technology for protein structure and design. This platform includes three key components: ESMC, a language model trained on 2.8 billion protein sequences; ESMFold2, which rapidly transforms these sequences into atomic-resolution 3D structures; and ESM Atlas, a comprehensive database organizing vast amounts of protein data. This announcement is significant as it provides researchers with unprecedented access to tools that can accelerate drug discovery, particularly in cancer and immunology, where designing effective protein binders is a major challenge. The technological implications of this release are profound. ESMFold2 can produce functional protein binders in days rather than years, a transformative shift from empirical to computational design, which is crucial for developing therapies targeting diseases. With successful outcomes in lab validations against key cancer-related proteins, this platform not only enhances the efficiency of therapeutic development but also democratizes access to cutting-edge AI tools, fostering a collaborative scientific ecosystem. Biohub's commitment to open science aims to expedite personalized medicine efforts by enabling researchers to harness AI advancements for deeper biological understanding and targeted treatment solutions.