Biohub, a nonprofit research institute, said on Wednesday, Oct. 7, that it, the U.S. Department of Energy, the National Institutes of Health and new partners are putting $1.8 billion in funding, data, computing and new lab tools into open biological data built for AI. Biohub calls it the largest coordinated commitment to generating AI-ready biological data to date. The aim is AI that can predict how a living cell responds when something changes, so scientists can try some experiments on a computer instead of only in the lab.
Scientists call that goal a virtual cell. "An accurate predictive model of biology could dramatically accelerate scientific discovery by enabling scientists to perform experiments digitally," said Alex Rives, Biohub's head of science. The obstacle is data. AI models learn from examples, and Biohub said in April that the goal needs vastly more data than exists today. According to a news report, Rives said today's cell datasets cover hundreds of millions of cells, while an accurate model will need billions and eventually trillions.
The Department of Energy will invest more than $500 million over five years in lab measurement, modeling and computing, through its Genesis Mission, a cross-agency effort the department leads. The National Institutes of Health will contribute datasets built with more than $500 million in earlier federal funding, which Biohub will put into a common format for training AI. Google DeepMind, Isomorphic Labs, a drug discovery company, and Meta are together investing $300 million. According to a news report, Biohub is a philanthropic venture of Meta's chief executive, Mark Zuckerberg, and Priscilla Chan. That adds to the $500 million Biohub committed when it launched the effort, called the Virtual Biology Initiative, in April. So about $1 billion of the total was not new on Wednesday: it is the value of federal data already paid for, plus Biohub's April pledge.
Biohub says the result will be an open resource for researchers. According to a news report, though, Rives said companies that fund the work get a head start, a period to work on the data before it becomes public, while the government-funded work carries no such restriction. Research groups including the Allen Institute, the Broad Institute, the Gladstone Institutes, the Human Cell Atlas, the Human Protein Atlas and the Wellcome Sanger Institute are taking part, and NVIDIA is supporting the computing.
No accurate virtual cell exists yet. According to a news report, Rives expects a first dataset in about a year and accurate predictive models within five years. Biohub has not said how long the companies' head start lasts or when the first data will be public.