DeepMind’s AI has modeled 20,000 human protein structures. In total, AI the algorithm has modeled 350,000 protein structures.
Specialists of the AI laboratory DeepMind have collected the most complete database of human protein structures created by the AlphaFold neural network. The set contains 20,000 3D models.
The first version of the algorithm AI predicted the protein structure with an accuracy of 98.5%, and also calculated the position of about 60% of the amino acids in these chains. Scientists hope that this will help build a proteomic map that includes all proteins encoded by the human genome.
According to the researchers, AlphaFold outperforms the particle accelerator and other instruments used to study the structure of proteins over the past 50 years. In particular, scientists were able to accurately determine the structure of molecules associated with diabetes, Wolfram syndrome and other diseases that have not previously been possible to identify experimentally.
In addition to human proteins, the researchers performed calculations with AI for two dozen living organisms, including fruit flies, mice and yeast. In total, AI the algorithm has modeled 350,000 protein structures.
In the near future, scientists plan to expand the base and determine the three-dimensional structure of 100 million proteins known to science. They hope that this will bring humanity closer to understanding the structure of living organisms and the invention of new types of drugs.