Image: © sergunt/Stock.adobe.com
The developers of DeepMind have claimed that their latest AI program has solved the age-old mystery of protein folding.
The AI developed by Google-owned DeepMind may have previously dominated in board games such as Go and video games such as Starcraft II, but now it has helped solve a biological mystery that has perplexed scientists for 50 years.
Writing in a blog post, the company said its latest AI program, AlphaFold, has solved the problem of ‘protein folding’, which the organisers of the biennial Critical Assessment of protein Structure Prediction have tasked others with finding a solution to.
Publishing its findings in a paper still awaiting peer review, DeepMind said that AlphaFold can accurately predict how proteins fold into 3D shapes. Because a protein’s shape is closely linked with its function, the ability to predict its structure unlocks a greater understanding of what it does and how it works.
Existing techniques to examine protein structures typically use nuclear magnetic resonance and X-ray crystallography – in addition to newer methods such as cryo-electron microscopy – that depend largely on time-consuming trial and error.
Based on the results seen with AlphaFold so far, DeepMind claimed that its AI may prove especially helpful for important classes of proteins, such as membrane proteins, that are very difficult to crystallise and therefore challenging to experimentally determine.
With this new knowledge of protein folding, DeepMind said it could escalate efforts to develop treatments for diseases or find enzymes that break down industrial waste. AlphaFold was trained on publicly available data on approximately 170,000 protein structures together with large databases containing protein sequences of unknown structure.
Using a relatively modest amount of computing power compared with most large state-of-the-art systems used in machine learning, the training took just a couple of weeks.
‘A stunning advance’
Commenting on the breakthrough, Prof Venki Ramakrishnan, a Nobel laureate and president of the Royal Society, said: “This computational work represents a stunning advance on the protein-folding problem, a 50-year-old grand challenge in biology.
“It has occurred decades before many people in the field would have predicted. It will be exciting to see the many ways in which it will fundamentally change biological research.”
DeepMind said that in addition to publishing a peer-reviewed paper on AlphaFold, it will now look to find a way to provide greater access to the system in a scalable way. It also aims to see how protein structure prediction could be useful in future pandemic response efforts.
“AlphaFold is one of our most significant advances to date but, as with all scientific research, there are still many questions to answer,” DeepMind said.
“Not every structure we predict will be perfect. There’s still much to learn, including how multiple proteins form complexes, how they interact with DNA, RNA, or small molecules, and how we can determine the precise location of all amino acid side chains.”
