Google, Meta and US Government Back $1.8 Billion AI Biology Initiative
A major international effort to combine artificial intelligence with biological research has expanded into a $1.8 billion initiative involving the US government, Google DeepMind, Meta and other scientific organizations.
The project, known as the Virtual Biology Initiative, aims to generate large-scale biological datasets that can be used to train AI systems capable of predicting how living cells behave.
The long-term objective is to create increasingly accurate digital models of biology that could help scientists understand diseases, test potential treatments and accelerate the development of new medicines.
What Is the Virtual Biology Initiative?
The Virtual Biology Initiative is an effort led by Chan Zuckerberg Biohub to create an open, standardized biological data resource for artificial intelligence research.
The initiative focuses on collecting information about how cells respond to different conditions and interventions.
Scientists could eventually use these datasets to train AI models that predict what may happen to a cell when it is exposed to a drug, genetic change or other biological intervention.
Instead of performing every experiment physically, researchers could potentially use predictive AI models to identify promising experiments before testing them in laboratories.
Nearly $2 Billion Is Being Committed
The expanded initiative represents approximately $1.8 billion in combined funding, data, computing resources and measurement technology.
Biohub's original commitment to the project is $500 million.
The US Department of Energy is contributing more than $500 million over five years toward biological research, measurements, AI analysis, modeling and computation.
The National Institutes of Health will contribute existing biomedical datasets, repositories and knowledge resources supported by more than $500 million in previous federal investment.
Google DeepMind, Isomorphic Labs and Meta are collectively investing another $300 million in the initiative.
Google and Meta Enter the Virtual Cell Race
The involvement of Google and Meta adds major artificial intelligence expertise to the project.
Google DeepMind has already developed AI systems for biological research, while Isomorphic Labs focuses on using AI for drug discovery.
Meta has also invested heavily in AI research and large-scale computing infrastructure.
The combination of AI expertise, biological datasets and government scientific infrastructure could create a powerful research environment for developing predictive models of biology.
The Goal Is a Digital Model of Human Biology
One of the most ambitious goals is to develop a virtual cell.
A virtual cell would be an AI-based model capable of representing biological processes and predicting how cells respond to different conditions.
Such a system would not simply describe biological information. It would aim to simulate or predict biological responses using large quantities of experimental data.
If successful, scientists could use these models to explore biological questions digitally before conducting physical experiments.
Why Biological Data Is So Important
Artificial intelligence systems need large quantities of high-quality data to learn patterns.
AI has benefited enormously from the huge amount of text, images, video and other information available on the internet.
Biology presents a different challenge.
Researchers still do not have a comprehensive dataset describing how every type of human cell behaves under different conditions.
Biological systems are extremely complex, and cellular responses can change depending on genetics, environment, disease state, drugs and interactions with other cells.
The new initiative is designed to fill some of these gaps by generating biological measurements at much larger scale.
Advanced Microscopes and Measurement Technologies
Biohub says the initiative will use advanced technologies to capture biological information that has previously been difficult or expensive to measure.
One important technology is cryo-electron tomography, which can provide detailed information about structures inside cells.
The initiative will also use advanced microscopy and other technologies capable of measuring biological processes across large numbers of cells.
These technologies will generate the raw scientific information required to train future AI models.
AI Could Change Drug Discovery
Drug development is traditionally a long and expensive process.
Researchers must identify promising molecules, test their biological effects, conduct laboratory experiments and eventually evaluate potential treatments through clinical studies.
AI models capable of predicting biological responses could help researchers identify promising candidates earlier in the process.
Instead of testing every possibility experimentally, scientists could use computational models to narrow down the most promising options before moving into laboratory testing.
This does not eliminate the need for physical experiments or clinical trials, but it could potentially make parts of the research process more efficient.
Open Data Is a Major Part of the Project
The initiative is designed around the creation of an open biological data resource that can eventually be used by researchers around the world.
Biohub says the project will work toward shared standards, common identifiers and a unified access system so that datasets generated by different organizations can work together.
This is important because biological research is currently spread across universities, government institutions, research organizations and private companies.
Creating standardized datasets could make it easier for researchers to train and compare AI models.
NVIDIA Is Also Supporting the Initiative
NVIDIA is supporting the Virtual Biology Initiative with accelerated computing infrastructure, specialized software and technical expertise.
Large biological datasets and AI models require substantial computing power, particularly when researchers are analyzing images, molecular structures and complex cellular interactions.
Access to high-performance computing could therefore become an important component of the project's development.
Scientists From Around the World Are Being Brought Together
The project involves organizations and scientific communities with experience in large international research collaborations.
Groups including the Allen Institute, Broad Institute, Gladstone Institutes, Human Cell Atlas, Human Protein Atlas and Wellcome Sanger Institute are participating in the wider effort.
The goal is to combine expertise from AI, biology, medicine, imaging, data science and computational research.
The Five-Year Ambition
Biohub's broader vision is to dramatically accelerate the pace of biological discovery.
The organization believes predictive models could allow scientists to perform some experiments digitally, helping researchers explore hypotheses before committing resources to laboratory work.
The initiative is designed to expand the amount of biological data available to AI models and develop technologies capable of processing that information at much larger scale.
The first major datasets are expected to become available as the program develops, with the long-term objective of producing increasingly capable predictive models.
Building a Virtual Cell Is Extremely Difficult
Despite the enormous investment, scientists caution that accurately modeling biology remains one of the hardest problems in science.
Human cells contain millions of interacting components, and biological processes operate across different time and spatial scales.
A useful virtual cell would therefore require far more than simply collecting large datasets.
Researchers will need better measurements, better AI architectures, more computing power and experimental validation to determine whether predictions accurately reflect real biological behavior.
A New Era for AI and Science
The $1.8 billion initiative demonstrates how artificial intelligence is expanding beyond language, images and software development into fundamental scientific research.
Instead of using AI only to analyze existing scientific information, researchers are now attempting to create entirely new datasets designed specifically for training scientific AI models.
If successful, the approach could change how scientists study disease and develop medicines.
What Comes Next?
The Virtual Biology Initiative will focus on generating biological measurements, organizing datasets and developing the computational infrastructure required to train predictive AI models.
The project will also need researchers around the world to adopt common standards and contribute complementary biological information.
The ultimate ambition is to create AI systems capable of predicting how cells respond to changes with enough accuracy to guide real scientific experiments.
A fully functioning virtual cell remains a major scientific challenge, but the latest $1.8 billion commitment marks one of the largest coordinated efforts yet to combine artificial intelligence, biological data and high-performance computing.
If the project succeeds, the next generation of AI may not simply answer questions about biology. It could help scientists predict what living cells will do before the experiment is performed.
Journalist: Vijay Singh