DeepMind says AlphaFold has predicted the structure of almost every protein cataloged by science, over 200M in total, a problem in biology for decades
Artificial intelligence firm DeepMind has transformed biology by predicting the structure of nearly all proteins known to science in just 18 months …
New ScientistMatthew Sparkes
Context & Ripple Effects
AlphaFold’s path ran from a competition-winning protein-structure model in 2018 to DeepMind’s 2020 claim that it had cut a key folding task from months to hours. This report marks the shift from demonstrating accuracy on a hard scientific problem to operating at catalog scale.
That scale also establishes the base for AlphaFold’s later expansion into predictions for molecules beyond proteins, including ligands and nucleic acids. The significance is less a single biological result than a reusable computational capability across molecular research.
First-order effects
DeepMind moves AlphaFold from a protein-folding breakthrough to coverage across more than 200 million cataloged proteins, making breadth of prediction a central part of the project’s value.
Researchers working on cataloged proteins can orient work around a predicted structural starting point rather than treating each structure solely as a long-running folding problem.
Second-order effects
Drug-discovery teams gain a larger pool of protein targets for structure-informed investigation, consistent with AlphaFold’s reported success in work on sleeping-sickness treatments.
DeepMind’s next competitive task becomes extending prediction beyond proteins; its later model’s coverage of ligands and nucleic acids shows the response was to broaden molecular scope rather than stop at the protein catalog.
Third-order effects
If molecular prediction continues to expand from proteins to interacting molecules, AI’s role in biology shifts from solving discrete benchmark problems toward a general-purpose research layer for chemistry and therapeutics.
The advantage increasingly lies with labs that can turn model outputs into usable scientific workflows, not merely with those that demonstrate protein-structure accuracy in competitions.
The trend: AlphaFold is part of a shift from AI scientific benchmarks toward broad molecular prediction systems that can support downstream research workflows.
Today in partnership with @emblebi, we're releasing predicted structures for nearly all catalogued proteins known to science, which will expand the #AlphaFold database by over 200x - from nearly 1 million to 200+ million structures: https://dpmd.ai/... 1/ https://twitter.com/...
DeepMind and EMBL-EBI have released the predicted structures of over 200 million proteins covering almost every organism that has had its genome sequenced. These are now openly available to the scientific community via the AlphaFold Database. https://www.embl.org/... https://twit…
Momentous AI news from @DeepMind & @emblebi that could greatly advance our knowledge of biology. “Predicting the structure of almost every known protein is a UK-led accomplishment on par with anything else in the country's long history of scientific discoveries.” - @DameWendyDBE …
A monumental milestone that will accelerate our understanding of disease and discovery of new medicine. The structures of the molecular machinery predicted by @DeepMind will be key to unlocking longer and healthier lives 🧬 https://twitter.com/...
Since its launch last year, the @DeepMind and @emblebi AlphaFold protein database has helped scientists tackle complex, global problems. This announcement today is not only another huge advancement, but a step towards insuring the world is prepared for future pandemic threats. ht…
“Hassabis said his dream is that AI could not just help figure out the structure of proteins, but become a ‘significant part of the discovery process for new drugs and cures.’” https://www.technologyreview.com/ ...
One year after open-sourcing its AlphaFold Protein Structure Database, @DeepMind and @emblebi are expanding it from nearly 1M to over 200M protein structures, covering almost every organism that has had its genome sequenced, a huge milestone. https://twitter.com/...
The archive is a “gift to humanity” according to @ewanbirney. And Keith Willison at @imperialcollege says people are joking that those working in the area are going to be unemployed. Cracking a SINGLE protein once took Keith 8 years. 5/9 https://twitter.com/...
In 2020 @DeepMind used AI to predict - with high accuracy - the structure of any given protein in minutes. By 2021 they'd unpicked almost every protein found in the human body. https://www.newscientist.com/ ... 3/9
Proteins are the building blocks of life. They're in our bones, blood and brain. Determining how they fold up and tangle is key to developing new drugs. It's also REALLY hard. It takes researchers literally years to untangle a single protein. https://www.newscientist.com/ ... 2/9
The impact will be huge, and @ewanbirney sums it up well: “I've seen many of these moments where you can sense the landscape shifting under you and the provision of new resources, and this has been one of the fastest. Two years ago we didn't realise that this was feasible.” 9/9
Without looking, I'm going to bet one of the proteins I've been working on for years isn't among the structures it's been able to predict... https://twitter.com/...
“DeepMind said it had excluded viruses from the database to prevent this data from being potentially weaponised by bad actors or bioterrorists.” Love to rely on the kindness/foresight of tech giants to mitigate existential risks https://www.ft.com/...
And here we go! Jaw-dropping result from DeepMind, essentially solving an entire field of science in just 18 months. Possibly the biggest result in AI, ever? https://www.newscientist.com/ ...