The release of predicted protein structures by the AI company DeepMind has made a significant impact in the scientific community. DeepMind’s previous breakthrough in 2021 involved the release of predicted structures for approximately 350,000 proteins. However, the company has now gone even further by unveiling the likely structures of over 200 million known proteins from various organisms. This achievement demonstrates the potential of AI in facilitating drug development and evolutionary studies.
DeepMind’s founder and CEO, Demis Hassabis, announced the release of structures for the entire protein universe during a press conference in London. This extensive collection of protein structures is a result of AlphaFold, one of the AI programs that successfully addresses the long-standing challenge of protein folding. AlphaFold’s newly predicted structures were integrated into an existing database through a collaboration with the European Molecular Biology Laboratory’s European Bioinformatics Institute (EMBL-EBI). This collaboration has empowered structural biologists to easily access protein structures, similar to conducting a keyword search on Google.
Eric Topol, director of the Scripps Research Translational Institute, expressed awe at the monumental advance in life science achieved by AlphaFold. He highlighted the potential of these new protein structures to unravel biological mysteries on a daily basis. Ewan Birney, deputy director general of EMBL, described DeepMind’s structure release as remarkable and emphasized the implications it would have on global researchers’ experiments.
AlphaFold’s protein predictions encompass diverse organisms, from bacteria to plants to vertebrates, including mice, zebrafish, and humans. Each protein prediction was generated by AlphaFold in approximately 10 to 20 seconds. To effectively present the colossal number of structures in the database, DeepMind collaborated closely with EMBL-EBI.
Since its launch last year, more than 500,000 researchers have utilized the database provided by DeepMind. Hassabis anticipates a new era in digital biology, in which AI-predicted protein structures can aid in designing small molecules that target specific proteins relevant to medical conditions, thereby offering potential treatments. Additionally, scientists are utilizing these structure predictions to develop vaccine candidates, investigate fundamental biology questions, and explore the evolution of proteins during the origins of life.
However, Hassabis emphasized that the release of these structures is just the beginning and that there is still much more biological and chemical research to be conducted.