The award-winning team behind Google DeepMind's AlphaFold program is no more, according to Financial Times. Google has reassigned most of the team's key members and the original authors of its papers, while a few others have already left the company. AlphaFold is an AI program that can accurately predict three-dimensional structures of proteins from their amino acid sequences in minutes instead of years. It's now being used to accelerate drug discovery, develop vaccines and understand the structural changes in proteins associated with neurodegenerative diseases like Alzheimer's and Parkinson's.
DeepMind started developing AlphaFold in 2018. In 2020, it was recognized as a solution to humanity's 50-year-old "protein folding problem," which sought to answer how amino acids automatically fold into complex 3D shapes. Those shapes determine the biological role of a protein. Scientists had identified the structures of roughly 170,000 proteins over the past 50 years, using tools and techniques like X-ray and nuclear magnetic resonance. The AlphaFold team took information from those previous work and then fed it to their AI to train AlphaFold.
In 2021, Nature published the papers with AlphaFold's methodology and the structure predictions of the entire human proteome, or the complete set of proteins expressed by our species. DeepMind then launched the AlphaFold Protein Structure Database, giving researchers free access to over 200 million protein structure predictions. In 2024, DeepMind CEO Demis Hassabis and John Jumper, who was a staff research scientist when the project began and who eventually became a VP and engineering fellow, won the Nobel Prize in Chemistry for their work on AlphaFold.
Keep the slop, kill the science. Got it.
This is like saying computational biology is not suited for real science, as AlphaFold beat existing docking search algorithms on accuracy. Neither means you just get an answer spit out and your science is done, but it doesn't mean they're not productive tools in the process.
And all of these predictive methods, ML and not, are biased to existing data, because that's all the information we have to guide their development and gauge their success.