Mark Zuckerberg and Priscilla Chan's Biohub has teamed up with Meta, Google DeepMind, Isomorphic Labs, and the U.S. government to create what they call a "universal virtual cell." They plan to train AI to predict how a living cell will behave under different conditions, thus testing some scientific hypotheses before they even go into the lab.
Mark Zuckerberg and Priscilla Chan's Biohub has teamed up with Meta, Google DeepMind, Isomorphic Labs, and the U.S. government to create what they call a "universal virtual cell." They plan to train AI to predict how a living cell will behave under different conditions, thus testing some scientific hypotheses before they even go into the lab.
This is reported by Axios.
AI already works quite well with individual biological structures, such as proteins. But modeling an entire living cell is much more difficult, because it consists of a huge number of interconnected processes, and data about them is still insufficient.
So, in the first phase, researchers will create a large-scale map of cell biology: they will collect data about cells and how they respond to various changes. In the future, the AI model should allow scientists to ask questions like: “What will happen to the cell if we change this molecular mechanism?” and get a prediction on the computer. This should help weed out less promising experiments and spend time and money in the laboratory only on those that can really provide new knowledge.
In the future, such models could be used to study aging and regeneration, as well as to find the molecular causes of diseases. In an even more ambitious scenario, the company could create a model that can analyze a specific person's disease, predict its molecular causes, and suggest which intervention might work best.
But one of the main obstacles is the lack of the necessary data. Models can be trained on huge arrays of already existing texts, but information about how living cells behave still needs to be obtained through real-world measurements and experiments in many cases.
Previously, dev.ua wrote that Stanford Medicine researchers created Virtual Biotech, a virtual biotechnology company with 37,000 AI agents.
