Biohub launches AI world model transforms protein biology accelerates drug discovery in United States California research labs

Biohub, a research initiative supported by Mark Zuckerberg and Priscilla Chan, has unveiled a new AI-driven "world model" designed to understand and engineer proteins at scale. Utilizing fourth-generation evolutionary scale modelling (ESM), the system learns from massive datasets of protein sequences shaped by billions of years of evolution. This allows researchers to predict protein structures and design entirely new molecules, potentially revolutionizing how the building blocks of life are manipulated for medical use.

Early experimental results have shown significant promise in oncology and immunology, with AI-designed protein binders successfully reactingivating immune cells in laboratory tests. To foster global scientific advancement, Biohub has committed to an open-source strategy. The models will be accessible through platforms such as AWS Bio Discovery and SandboxAQ, and researchers will be provided with compute credits to lower the barrier to entry for labs with limited computing budgets.

While the technology promises to significantly shorten drug discovery timelines and democratize access to advanced design tools, it also necessitates rigorous validation. Experts emphasize that AI-designed biological systems must undergo strict oversight to manage safety concerns and unintended biological effects. Ultimately, this shift marks a transition from reactive medicine to a predictive, design-driven era of biology.

(Source：Event Coverage)