DeepMind’s AI researchers have assembled the most comprehensive database of human protein structures, created by the AlphaFold neural network. The dataset comprises 20,000 three-dimensional models.
Today with @emblebi, we’re launching the #AlphaFold Protein Structure Database, which offers the most complete and accurate picture of the human proteome, doubling humanity’s accumulated knowledge of high-accuracy human protein structures — for free: https://t.co/vtBGmTkKhy 1/ pic.twitter.com/XgBQTn2fuC
— DeepMind (@DeepMind) July 22, 2021
The first version of the algorithm predicted a protein’s structure with an accuracy of 98.5%, and estimated the position of around 60% of the amino acids in these chains. Scientists hope this will help build a proteome map that includes all proteins encoded by the genome.
According to the researchers, AlphaFold outstrips particle accelerators and other instruments used to study protein structures over the past 50 years. In particular, scientists were able to determine the precise structures of molecules linked to diabetes, Wolfram syndrome and other diseases that previously could not be identified experimentally.
In addition to human proteins, the researchers conducted calculations for two dozen living organisms, including fruit flies, mice and yeast. In total, the algorithm modelled 350,000 protein structures.
In the near future, scientists plan to expand the database and determine the three-dimensional structure of 100 million proteins known to science. They hope this will bring humanity closer to understanding how living organisms are put together and to the invention of new types of medicines.
In March, scientists from NVIDIA and Harvard University developed artificial intelligence that speeds up the analysis of the human genome.
In July, researchers from the University of California, San Francisco presented a neural interface that enabled a paralyzed man to speak.
Follow ForkLog news in Telegram: ForkLog AI — all the news from the AI world!