Driving the news: Biohub, the Department of Energy, the National Institutes of Health, Google DeepMind, Isomorphic Labs, Meta and a range of scientific organizations are collaborating to create and standardize data for what Biohub calls a “universal virtual cell.”
The big picture: The goal is to use AI to virtually explore many more scientific questions in biology, allowing scientists to reserve expensive lab work for the experiments most likely to teach them something important.
- “If we can put more and more reasoning and intelligence into every question we actually ask in the lab, the value of these empirical results will be much greater,” Alex Rives, Biohub’s chief scientific officer, told Axios.
Much of the recent advances in AI is the result of combining better algorithms, more computing power and huge amounts of data.
- Biology presents an additional challenge: Most of the information AI needs does not yet exist and must be carefully measured from the physical world.
- “We are at the beginning of a new scientific paradigm with AI,” Rives said.
Yes, but: Biology is more difficult than many other areas of AI because researchers need what Rives calls “empirical AI”: models that learn from biological evidence and can accurately predict what happens in the physical world.
- “The big challenge in biology is to bridge the gap between the computational and digital world and the real physical world of biology and life,” Rives said. “The way to do that is through data.”
- Such models could eventually help scientists study fundamental questions such as how aging and regeneration work, or medical questions such as the molecular mechanisms responsible for Alzheimer’s disease.
How it works: The first phase will create a broad map of cell biology, bringing together different types of information about cells and how they respond to changes.
- Further, Rives envisions models that could examine an individual’s disease and predict its molecular causes and how best to intervene.
The plot: Business partners will have one year of exclusive access to the data they develop before it is shared publicly.
- Rives said this temporary benefit is intended to give companies a reason to contribute financially while ensuring that the data obtained becomes an open scientific resource.
- “We need to incentivize commercial players to participate in this, and the embargo period creates that,” Rives said. “But it’s a one-year embargo. So that means that very quickly, the data becomes widely available to scientific efforts.”
Zoom: This effort involves $1.8 billion in funding, data, computing and measurement technologies.
- DOE plans to invest more than $500 million over five years in biological measurement, modeling and computing, while NIH contributes data sets and other resources resulting from more than $500 million in previous federal investments.
- Google DeepMind, Isomorphic Labs and Meta are collectively investing an additional $300 million, while Biohub previously committed $500 million to the effort.
Flashback: When Biohub announced its initial $500 million effort in April, Rives told Axios that one of the biggest unanswered questions was whether cell biology would exhibit the same kind of “scaling laws” seen elsewhere in AI — with models getting better and better as they are trained on increasing amounts of data.
What we’re looking at: Rives thinks researchers won’t have to wait very long to find out if the bet works.
- He said less than a year after obtaining the first large-scale data set, researchers should be able to train models, measure their capabilities, and determine what types of additional biological data improve them.
- “It’s worked in every field, and it works in biology as well,” Rives said, highlighting AI’s advances in protein biology.
Gn bussni
