Understanding Proteins

Alphabet (the parent company of Google) recently announced the development of an artificial intelligence (AI) algorithm that can predict the shape of a protein from its component amino acids. Called AlphaFold, the program uses deep learning to model a protein’s structure in a matter of days with a high level of accuracy.

Proteins are made up of twenty different amino acids that can assemble in as many as 10300 different shapes. More than a simple chain of molecules, proteins form complex “folds” based on the interaction of these amino acids. Each shape determines the function of that protein in sustaining life; for example, enzymes to maintain body chemistry, myosin to keep muscles working and hemoglobin to carry oxygen in the blood. While there are more than 200 million known proteins, the structure of only 170,000 have been identified. This has been done through experimental trial and error using expensive, specialized equipment and techniques like nuclear magnetic resonance, x-ray crystallography and cryo-electron microscopy, all of which can take years.

The AlphaFold algorithm was “trained” using what is already known about these 170,000 proteins to piece together protein structures based on a simple list of components. The breakthrough could have profound implications in the medical arena to increase understanding of diseases and potential treatments, as well as environmental applications in the development of new enzymes for breaking down industrial waste and other pollutants.