What is the story about?
A philanthropic venture founded by Mark Zuckerberg and his wife, Priscilla Chan, launched a new “world model” for protein biology on Wednesday, aiming to accelerate drug discovery through artificial intelligence.
Proteins are the body’s essential molecular machinery, performing diverse roles ranging from building cellular structures to generating energy. Pharmaceutical companies are increasingly turning to AI to strengthen research and development efforts, betting heavily on advanced modelling tools to speed up discoveries.
The newly introduced world model consists of open-source AI systems designed to improve scientists’ understanding of proteins and enhance their ability to design them. Researchers have already used these AI models to create new protein binders targeting cancer and immune-related diseases. In laboratory tests, these binders were able to reactivate immune cells successfully.
The models will also be made available through platforms including AWS Bio Discovery and SandboxAQ. Founded in 2015, the Chan Zuckerberg Initiative consolidated its biomedical research work under Biohub in November 2025, including the acquisition of AI biology startup EvolutionaryScale.
Biohub said the system could compress years of protein research into hours or days, potentially accelerating disease research and treatment development. According to the organisation, the release includes a protein structure prediction model, a protein language model, and the ESM Atlas — a massive database containing 6.8 billion proteins and 1.1 billion predicted protein structures.
Together, these tools could create a faster and more streamlined way of identifying diseases and developing treatments. However, experts caution that the technology is still several steps away from producing drugs capable of passing clinical trials, and any therapeutic applications would still require extensive safety testing.
Biohub believes AI could eventually make biology more programmable, allowing scientists to test ideas computationally before moving them into laboratories for physical experimentation.
The organisation says it aims to provide researchers worldwide with an open discovery engine for protein structure prediction, protein design, and biological discovery. Reports suggest Biohub researchers used the model to design protein binders against five major cancer and immunology targets. Laboratory experiments showed hit rates of 36–88% for compact mini-binders and 15–29% for antibody-derived formats with confirmed binding success.
Such rapid advances could significantly reduce the time needed for disease discovery and treatment selection.
Biohub is also being described as a trained evolutionary record of life itself — encompassing billions of protein sequences spanning organisms ranging from deep-soil bacteria to life forms thriving in extreme environments. Its training objective is relatively straightforward: predict the amino acids selected through evolution, since evolution naturally preserves proteins best suited for survival and function.
Patterns preserved across billions of years of biological data implicitly encode the physical rules governing protein function. Biohub ultimately hopes to achieve cures and prevent diseases by developing a deeper understanding of biology and ensuring that the tools enabling that understanding remain accessible to scientists everywhere.
(With Inputs from Reuters)
Proteins are the body’s essential molecular machinery, performing diverse roles ranging from building cellular structures to generating energy. Pharmaceutical companies are increasingly turning to AI to strengthen research and development efforts, betting heavily on advanced modelling tools to speed up discoveries.
Streamlining medical treatment
The newly introduced world model consists of open-source AI systems designed to improve scientists’ understanding of proteins and enhance their ability to design them. Researchers have already used these AI models to create new protein binders targeting cancer and immune-related diseases. In laboratory tests, these binders were able to reactivate immune cells successfully.
The models will also be made available through platforms including AWS Bio Discovery and SandboxAQ. Founded in 2015, the Chan Zuckerberg Initiative consolidated its biomedical research work under Biohub in November 2025, including the acquisition of AI biology startup EvolutionaryScale.
Biohub said the system could compress years of protein research into hours or days, potentially accelerating disease research and treatment development. According to the organisation, the release includes a protein structure prediction model, a protein language model, and the ESM Atlas — a massive database containing 6.8 billion proteins and 1.1 billion predicted protein structures.
Together, these tools could create a faster and more streamlined way of identifying diseases and developing treatments. However, experts caution that the technology is still several steps away from producing drugs capable of passing clinical trials, and any therapeutic applications would still require extensive safety testing.
Biohub believes AI could eventually make biology more programmable, allowing scientists to test ideas computationally before moving them into laboratories for physical experimentation.
Open discovery engine for researchers worldwide
The organisation says it aims to provide researchers worldwide with an open discovery engine for protein structure prediction, protein design, and biological discovery. Reports suggest Biohub researchers used the model to design protein binders against five major cancer and immunology targets. Laboratory experiments showed hit rates of 36–88% for compact mini-binders and 15–29% for antibody-derived formats with confirmed binding success.
Such rapid advances could significantly reduce the time needed for disease discovery and treatment selection.
Biohub is also being described as a trained evolutionary record of life itself — encompassing billions of protein sequences spanning organisms ranging from deep-soil bacteria to life forms thriving in extreme environments. Its training objective is relatively straightforward: predict the amino acids selected through evolution, since evolution naturally preserves proteins best suited for survival and function.
Patterns preserved across billions of years of biological data implicitly encode the physical rules governing protein function. Biohub ultimately hopes to achieve cures and prevent diseases by developing a deeper understanding of biology and ensuring that the tools enabling that understanding remain accessible to scientists everywhere.
(With Inputs from Reuters)
