Artificial intelligence (AI) is helping scientists tackle some of the world’s most challenging questions, including those related to microbiology and protein analysis.
Proteins are the building blocks of living organisms. They consist of strings of amino acids, which represent a specific class of organic compounds. Depending on the type of protein molecule that the body is trying to make, these strings then fold into specific shapes. The precise shape of these formations has puzzled scientists for decades. It often takes years of research for scientists to determine the mechanisms behind how a given protein folds and how it knows which type of protein to become.
AlphaFold, the AI from DeepMind, is an example of a company that’s attempting to crack this protein-folding code. With a high level of accuracy, AlphaFold predicts what shape the amino acid strings of proteins will transform into. AlphaFold’s AI was tested in the global competition called Critical Assessment of protein Structure Prediction, or CASP. During the competition, researchers from around the world came together to try and advance the methods of identifying proteins’ potential three-dimensional shapes based off of their amino acid sequences.
In the most recent competition, CASP put AlphaFold and others to the test in order to see how they held up against the traditional methods of human-based protein research. Competing AI systems vied to see who could most accurately predict a protein’s future shape. Judges then compared the accuracy of the AI systems to previous human-based research findings regarding protein shapes.
This year, AlphaFold broke a record. By a wide margin, their system made the most accurate predictions for any CASP participant in the competition’s 25-year history. While the data will still have to be peer-reviewed and published, this newfound AI capability is a huge advancement for science, specifically in terms of biological research. In comparison to the standard lab approach to solving they mysteries of protein folds, which can take years, AI systems like AlphaFold could greatly expedite this process, allowing for faster biological research and drug development in the future.
Watch a video of the AI system that predicts how proteins fold. Video clip here.