Chan Zuckerberg Biohub AI Predicts Structure of Over One Billion Proteins

Scientists at the Chan Zuckerberg Biohub have created an AI-powered model called ESMFold2 capable of predicting the structure of over one billion proteins. The open-source protein atlas surpasses Google DeepMind's previous solutions by hundreds of millions of entries and opens new possibilities for drug discovery and cancer treatment development.

Scientists at the Chan Zuckerberg Biohub have made a breakthrough in bioinformatics by creating an AI-powered protein atlas containing over one billion entries. The new model, named ESMFold2, can predict the spatial structure of proteins on an unprecedented scale.

ESMFold2 surpasses the current market leader—Google DeepMind's AlphaFold-based database—by hundreds of millions of entries. This means scientists now have access to a vastly broader selection of biological structures to study and use in drug development.

The protein atlas has been released as open-source code, meaning the global scientific community can access it freely and apply it in their research. The new tool is expected to have significant potential in accelerating precision medicine and cancer drug development, as it enables a better understanding of the molecular mechanisms of disease.

Predicting protein structure is one of the most complex tasks in the life sciences, because the three-dimensional shape of a protein determines its biological function. Accurate knowledge of protein structure is critical in drug design, as it enables the creation of compounds that bind to specific targets in cells within the body.