Google subsidiary DeepMind, an artificial intelligence (AI) research lab, has successfully predicted the structure of proteins… in 30 minutes. A speed record for a revolutionary paradigm shift.
To determine the three-dimensional structure of proteins, the experimental process is both long and expensive. London-based DeepMind, a Google company, has found an answer to this question with AlphaFold, a pioneering AI engine for determining the 3D shapes of proteins. In just a few minutes, this AI can successfully define structures with unprecedented accuracy. A result that previously required years of laboratory research. Out of a group of 100, AlphaFold determined the shape of about 60 proteins with levels of accuracy comparable to those of laboratory experiments, with the rest showing good results.
Alphabold was a standout performer at the 14th CASP (Critical Assessment of Structure Prediction), a bi-annual community-based experiment to determine and advance the state of the art in protein structure modeling based on amino acid sequence. Participants are invited to submit models for a set of proteins for which experimental structures are not yet public. Independent reviewers are then responsible for comparing the models with the experiment.
Revolutionizing the life sciences
In a press release, CASP14 organizer Dr. Andriy Kryshtafovych (University of California, Davis) deemed the achievement “truly remarkable.” A better understanding of protein structures coupled with the ability to predict them using AI will lead to a better understanding of life, health, and human disease. In the context of protein folding, structure determines function: the role of a given protein will therefore depend on its three-dimensional structure. Solving the mechanisms of protein folding is one of the major challenges of modern biology. This will allow us to understand certain pathologies and to develop new drugs. A huge leap forward for science with the availability of DeepMind’s work in open source, accessible to the greatest number of people!
