Small mutations in genes can lead to changes in a protein’s shape, meaning the protein can no longer fulfil its usual role in the body. In some cases, this can contribute to the development of cancers. The AI program, AlphaFold, has major implications for understanding and treating cancer. It can predict the shape of proteins, which can offer  clues as to how changes could affect function and lead to cancer. AlphaFold can also be used to find molecules that, because of their shape, may help develop cancer treatments.

It is relatively simple to find out the molecular formula of a protein, even though it may be tens of thousands of atoms long. It is however fiendishly difficult to predict how a protein might fold. It  has taken painstaking research using X-ray crystallography and nuclear magnetic resonance imaging, to generate vast quantities of data.

Today, DeepMind’s AlphaFold program, trained on known protein structures available through EMBL-EBI’s data resources, uses AI to predict millions more. DeepMind has partnered with EMBL to make millions of protein structure predictions available to researchers worldwide, giving us a 3D view of the known protein universe. This helps us understand health, disease and how life works at a molecular level.

Furthermore, the availability of predicted protein structures can accelerate experimental approaches to structure determination, which is still vital in understanding the details about how proteins interact with other molecules and how their shape changes in the process.