A boom in precise protein-structure prediction by AI

In more than 60 years since the first detailed structure of a protein was determined at atomic resolution by X-ray crystallography, a series of increasingly powerful experimental techniques had resulted, by 2020, in structural elucidation for well over a third of all proteins encoded by the human genome. Nevertheless, a large part of the proteins remained intractable for traditional laboratory methods, resulting in a major gap in our efforts to make sense of the protein-coding genome sequence information. 2021 has seen major advances in overcoming this limitation, based on AI-powered computational structure-prediction methods of unprecedented accuracy.In July, DeepMind’s second generation of the AlphaFold algorithm was used to generate a comprehensive atlas of protein structures for almost 99% of all human proteins, including tens of thousands of structures for critically important components of the human body that had evaded previous experimental characterisation. Excitingly, all this information is freely available to the global research community through the Alphafold Database hosted at the European Molecular Biology Laboratory in Cambridge.

This breakthrough was followed by another publication in August in which a group at the University of Washington in Seattle took AlphaFold’s AI approach to the next level. In a living cell, proteins rarely carry out their functions in isolation; instead, they engage in a complex molecular dance guided by protein-protein interactions. The enhanced algorithm managed to accurately predict the molecular details of these interactions, taking us an important step forward towards an understanding of the dynamics of human cell physiology.

Combined with the massive acceleration of genome sequencing, these new computational tools for predicting the detailed three-dimensional structure of the cellular machinery are rapidly being deployed in laboratories worldwide, enabling new strategies for drug discovery and making sense of the function of the human body. And the story of AI applications in biology is not going to end here. Watch this space in 2022. Eriko Takano

Eriko Takano is professor of synthetic biology at the University of Manchester