DeepMind, founded in 2010 and now part of Google, specializes in artificial intelligence (AI) programming. Recently, DeepMind achieved a breakthrough in protein structure elucidation, as researchers previously knew only 17 percent of amino acids (the smallest basic building blocks of proteins).
With the help of calculated structure analyses of the „AlphaFold“ program, it is possible to decode nearly 20,000 human protein molecules, i.e. almost the entire human proteome. The three-dimensional structure of proteins is of great importance, especially pharmaceutically. The 3D structure of a protein molecule determines its function and mechanism in the body of the individual organism.
The updated AI and machine learning program now has a 98.5 percent protein elucidation rate for humans and can predict protein structure for other organisms such as mice, corn and malaria mosquitoes. However, the accuracy of the predicted protein structures currently varies, so validation is needed.
DeepMind’s experts – including co-founder and CEO Demis Hassabis – believe they will be able to determine 130 million structures by the end of 2021. But the calculated and experimentally determined protein structures already match remarkably well, Hassabis explains.
The DeepMind researchers have made the „AlphaFold“ database publicly available so that other sciences can use the values and data determined using AI analysis for further research projects.
Source: www.pharmazeutische-zeitung.de