Scientists have developed a new artificial intelligence system DeepMind AlphaFold, capable of predicting the three-dimensional structure of a protein. The problem of protein folding (folding) has been recognized by biologists around the world as one of the greatest problems of modern science, and researchers have been struggling to solve it for about half a century. On scientific breakthrough reported on the DeepMind website.
The results of the Critical Assessment of Protein Structure Prediction (CASP), conducted by the scientific community since 1994, have shown that AlphaFold provides unprecedented accuracy in predicting protein structure. CASP test participants must present a system that can predict how a protein will fold from a sequence of amino acids, each with specific chemical properties (eg polarity and hydrophobicity). In this case, the forecast result is compared with the three-dimensional structure revealed experimentally.
For testing, only those proteins whose folding has been studied only recently or, in some cases, has not yet been studied are used. In the latter case, you need to wait until the three-dimensional structure is obtained for comparison.
The average score for AlphaFold was 92.4 on the Global Distance Test (GDT) metric. At the same time, the score of 90 GDT is considered competitive among the results obtained experimentally. This means that artificial intelligence is able in many cases to calculate the three-dimensional structure of proteins more accurately than using a number of laboratory methods.
The three-dimensional structure of proteins is closely related to their function. However, for each amino acid sequence, there are theoretically an astronomical number of folding methods, but only a few can be realized and play an important role for the body.