DeepMind, the AI technology company that is part of the Google parent alphabet, has made a major breakthrough in predicting AI-based protein structure. The company announced today that its AlphaFold System has officially resolved a major protein folding challenge that has baffled the scientific community for 50 years. The advancement of DeepMind’s AlphaFold capabilities could lead to significant advancement in areas such as our understanding of disease and the future discovery and development of drugs.
The test AlphaFold passed essentially shows that the AI can correctly figure out the structure of proteins with a very high degree of accuracy (actually accurate to the width of an atom) in just a few days – a very complex task that is critical to figuring out how best to treat illness and solving other major problems such as: B. to find out how ecologically hazardous material such as toxic waste can best be broken down. You may have heard of Folding @ Home, the program that allows users to bring their own computing power to home computers (and previously game consoles) into protein folding experiments. This massive global crowdsourcing effort was necessary because predicting portion folding using conventional methods takes years and is extremely expensive in terms of pure costs and computing resources.
DeepMind’s approach involves the use of an “attention-based neural network system” (basically a neural network that can focus on certain inputs to increase efficiency). It is able to continuously refine its own prediction diagram of possible protein folding results based on their folding history and deliver highly precise predictions as a result.
How proteins fold – or transition from a random series of amino acids as they originally formed to a complex 3D structure in its final stable form – is key to understanding disease transmission and how common diseases such as allergies work. Understanding the process of folding can potentially change it, stop the progress of an infection in the middle of the crotch, or, conversely, correct wrinkle errors that can lead to neurodegenerative and cognitive disorders.
DeepMind’s leap in technology could make accurately predicting these wrinkles much less time and resource intensive, which could dramatically change the pace at which our understanding of diseases and therapeutics is advancing. This could be useful to counter major global threats, including future potential pandemics such as the COVID-19 crisis that we are currently suffering, by predicting viral protein structures early with high accuracy as soon as new future threats such as SARS emerge. CoV-2, which accelerates the development of potentially effective treatments and vaccines.