DeepMind study shows how AI research can help solve a problem that has challenged scientists for decades
A study published in Nature has demonstrated how artificial intelligence (AI) research can drive and accelerate new scientific discoveries.
Alphabet subsidiary DeepMind has created a dedicated, interdisciplinary team to explore using AI to predict the 3D structure of a protein based solely on its genetic sequence.
Its system, AlphaFold, builds on prior research using large genomic datasets to predict protein structure.
Many diseases are linked to malformed protein, and predicting how chains of amino acids will fold into the 3D structure of a protein has challenged scientists for decades.
WHY IT MATTERS
The 3D models of proteins that AlphaFold generates are more accurate than those that have come before — marking significant progress on “the protein-folding problem”.
This could help researchers understand how the body works and design new cures for diseases such as Alzheimer’s, Parkinson’s, cystic fibrosis and Huntington’s — where misfolded proteins are believed to play a role.
THE LARGER TREND
DeepMind has been heavily involved in the healthcare space, using its AI platform for diagnosis within organisations including the NHS.
Another recent study published in Nature found that Google’s AI technology was more accurate in predicting breast cancer than mammograms.
Meanwhile a study published in Nature last year found that DeepMind’s Stream system was able to detect acute kidney injury (AKI) two days before it happens.
In August 2019, DeepMind’s strategic report revealed losses for the financial year of around $571 million (£470 million). Shortly afterwards, Google announced that the health team at DeepMind, was joining Google Health.
ON THE RECORD
Pushmeet Kohli, research lead for the DeepMind science team, said: “Our aim at DeepMind is to solve the few but key fundamental questions in science that could unlock the solution to thousands more. Protein folding is one such question — predicting the 3D structure of proteins could help in our understanding of the body and how it works, enabling scientists to design new, effective cures for diseases more efficiently.”
Demis Hassabis, CEO and co-founder of DeepMind, said: “These early signs of progress, which build on decades of work by many experts in the field, demonstrate the potential of AI to help us master fundamental scientific problems, and I look forward to seeing where this research goes next.”
David Jones, head of the UCL Bioinformatics Group, who advised the team on parts of the project, said: “Experimental techniques to determine protein structures are time consuming and expensive, so there’s a huge demand for better computer algorithms to calculate the structures of proteins directly from the gene sequences which encode them, and DeepMind’s work on applying AI to this long-standing problem in molecular biology is a definite advance.”