Mark Zuckerberg and Priscilla Chan today announced the launch of a first-of-its-kind initiative that combines artificial intelligence and frontier biology to dramatically accelerate scientific progress toward understanding and treating human disease.
Since its founding in 2016, Biohub’s interdisciplinary teams of scientists and engineers have developed breakthrough technologies for observing, measuring and programming biology at the cellular level. The agency has created the largest single-cell datasets and built large-scale computers dedicated to biological research—resources found nowhere else.
Today, advances in artificial intelligence are opening new frontiers for science. To meet this moment, Biohub is launching the first large-scale scientific initiative dedicated to advancing artificial intelligence for biology. The biomedical research organization is uniquely positioned to lead this effort, combining world-class computing capacity, cutting-edge artificial intelligence research and engineering, and advanced experimental and imaging technologies.
When we started, our goal was to help scientists cure or prevent all diseases this century. With advances in artificial intelligence, we now believe this may be possible much sooner. Accelerating science is the most positive impact we believe we can make. So we’re dealing with AI-powered biology for our next chapter.”
Mark Zuckerberg, co-founder of Biohub
To accelerate this work, EvolutionaryScale, an artificial intelligence research lab and utility company that has built pioneering large-scale artificial intelligence systems for the life sciences, will join the Biohub. Alex Rives, co-founder and chief scientist of EvolutionaryScale—and a computer scientist known for pioneering work pioneering the field of AI language modeling in biology—will serve as chief science officer, leading an integrated research strategy in experimental biology, data and artificial intelligence.
“Advances in artificial intelligence are already beginning to give us new tools for understanding and engineering biology,” said chief science officer Alex Rives. “As we bring together artificial intelligence, biological data generation at scale, and creative experimental science, it may be possible to significantly accelerate the rate at which fundamental new scientific discoveries can be made.”
The combined team of AI scientists, engineers and researchers will develop the datasets, laboratory technologies and modeling innovations needed to build the next generation of AI systems that will power the future of biological discovery and disease research. To fuel these efforts, Biohub is expanding its computing capacity tenfold, to 10,000 GPUs by 2028, and is channeling its resources into data generation and experimental biology.
Biohub is committed to four grand scientific challenges: developing a unified cell model based on artificial intelligence to predict and understand how cells behave inside the human body; advancing state-of-the-art imaging systems to visualize complex biological processes at an unprecedented scale; creating new instruments to monitor and regulate inflammation in real time; and the use of artificial intelligence to reprogram and harness the immune system for early disease detection, prevention and treatment. Together, these efforts aim to decipher the inner workings of human biology and enable scientific and medical breakthroughs.
“When I worked as a pediatrician at UCSF, I treated children with diseases whose conditions were, in many cases, still mysteries to science,” said Biohub co-founder Priscilla Chan. “What I wanted more than anything else was a way to see what was going on inside their cells—how genetic mutations were expressed in different cell types and what exactly was going down. Until now, that kind of understanding has been out of reach. Artificial intelligence is changing that.
A vision for AI-powered biology
Biohub’s AI-based biology initiative aims to build powerful AI systems that can reason and represent biology to accelerate science:
-
Virtual Biology: As we create high-fidelity digital representations of molecules, genomes, cells and living systems, we will increasingly be able to ask scientific questions digitally – conducting virtual experiments at a scale and pace far beyond what is possible in the laboratory. Digital representations of biology will enable scientists to simulate experimental outcomes and explore biological systems with unprecedented speed and depth, allowing them to ask and answer fundamental questions in ways never before possible.
-
Quick discovery: We are simultaneously working on scientific AI that can reason, learn, and synthesize the world’s scientific data to reveal fundamental insights and accelerate the rate at which new scientific discoveries can be made.
The Biohub will research, build and scale these powerful technologies – and make them available to scientists to drive key advances in biology and cutting-edge medical applications. As we make progress on these kinds of systems, we believe it may eventually be possible to make decades of discoveries in months.
We believe this will combine to unlock frontier medicine. Artificial intelligence could enable advances such as early disease detection, monitoring and prevention. Programmable cell drugs. personalized therapies based on gene editing; and approaches to the prevention and treatment of inflammatory and chronic diseases.
Early Milestones: Virtual Immune System and New Models of Artificial Intelligence
As Biohub builds towards this long-term roadmap, today we’re also sharing several updates about our work in progress:
Biohub launches Virtual Immune System, a landmark effort to model the enormous complexity of the human immune system. The goal is to change the understanding of immunological science – opening the door to engineering human health, simulating immunotherapies, reprogramming dysfunctional cells and preventing disease before it occurs.
And, Biohub has three new models on its virtual cell platform for free:
- VariantFormer: a model that directly translates personal genetic variants into tissue-specific patterns of gene activity.
-
CryoLens: an end-to-end, pretrained, large-scale model for cryoET that provides unsupervised structural similarity analysis. and
-
scLDM: a new AI model that can generate realistic single-cell data in silicon in unprecedented fidelity.
These complement other leading models recently made available, such as GREMLN, a graph-aware model for understanding how gene regulatory networks govern cell behavior, and rBio, a large-language conversational model designed to perform biological reasoning and make scientific knowledge more accessible, and other models, which provide a comprehensive platform for AI-based biology research.
Rapid advances in technology are driving a revolution in the life sciences—where frontier artificial intelligence and virtual biology are shaping this next era of biological discovery. Biohub, since its early days, will continue to engage responsibly with the global scientific community, and datasets and models will be freely shared to democratize access to cutting-edge technologies for research and discovery.
Source:
