For many proteins, however, “we are interested in understanding how their structure is altered by mutation and natural allelic variation, and this database will not address that,” AlQuraishi said. “Of course, the field is developing rapidly, so I hope that tools that can accurately model protein variants will appear soon,” he added.
For rare proteins with less available evolutionary information, the quality of AlphaFold’s predictions may also be less accurate, Peng said.
The move is the latest in DeepMind’s foray into “digital biology,” where “artificial intelligence and computational methods can help understand and simulate important biological processes,” Hassabis said. Hassabis also leads a new venture called Isomorphic Labs, also owned by Alphabet, that is developing artificial intelligence for drug discovery.
DeepMind’s head of scientific artificial intelligence, Pushmeet Kohli, said the company still faces many challenges in the life sciences, such as how proteins behave and how they interact with other proteins.
Hassabis said his dream is that AI could not only help figure out the structure of proteins, but also become “an essential part of the discovery process for new drugs and treatments.”