November 30, 2020

fifty-year-old science challenge solved with AI

Proteins form the machinery that keep all animals, plants and bacteria alive and well. Inside each of the cells that make up all living things, there are thousands of different proteins.  Understanding the complex shapes of these proteins is key to the study of life on earth, as well as deepening our understanding of disease and playing a pivotal role in developing cures.

Despite fifty years of work, scientists have struggled to find an accurate and efficient method to determine the shapes of proteins, but researchers now say an artificial intelligence solution devised by DeepMind has been found.  This has been presented for the first time at the Critical Assessment of Techniques for Protein Structure Prediction (CASP) conference on 30 November.

Journalists dialled in to this briefing to hear from DeepMind and CASP experts about this new solution to the protein folding problem and what this could mean for the future, explaining: 

• What do proteins do and why is their shape so important? 
• Why has it been so hard to determine the shapes of proteins?
• How has AI been used to solve the challenge and where does it take us now?

Speakers included:

Prof John Moult, Chair of the Community Wide Experiment on the Critical Assessment of Techniques for Protein Structure Prediction (CASP) from the University of Maryland

Dr Demis Hassabis FRS, CEO and Founder of DeepMind

Dr John Jumper, Senior Staff Research Scientist and AlphaFold Lead at DeepMind

Prof Dame Janet Thornton, European Bioinformatics Institute (EMBL-EBI)