Chan Zuckerberg Initiative’s rBio uses virtual cells to train AI, bypassing lab work
1 min read
Summary
Artificial intelligence (AI) has made an “important breakthrough” in cellular biology according to a recent announcement by the Chan Zuckerberg Initiative (CZI).
The breakthrough enables AI to learn using virtual simulations rather than the traditional method of conducting expensive laboratory experiments.
CZI believes the new way of learning could expedite drug discoveries and help researchers with hypotheses without having to conduct laboratory experiments.
The AI used for the learning is called rBio which is a large language model and is based on CZI’s own virtual cell model, TranscriptFormer.
rBio is able to interact with users in plain English and has been trained using reinforcement learning and virtual cell simulations.
The new model was tested against others that had been trained using experimental data and it performed just as well, if not better, according to rBio’s research paper.
CZI has said it will soon make rBio available to users through its Virtual Cell Platform.