Google DeepMind has developed a tool to predict the structure of proteins from their genetic sequence, marking a noteworthy example of using AI in the process of scientific discovery.

How it works: The system, called AlphaFold, models the complex folding patterns of long chains of amino acids, based on their chemical interactions, to form the three-dimensional shape of a protein. This is known as the “protein folding problem,” which has challenged scientists for decades.

Why it matters: The shape of a protein dictates its function in the body, so being able to predict a protein’s structure allows scientists to synthesize new protein-based drugs to treat diseases or new enzymes to break down pollutants in our environment.

Training data: The DeepMind team trained deep neural networks to predict the distances between pairs of amino acids and the angles between their chemical bonds, using the massive amounts of data available from genomic sequencing. The resulting system generates highly accurate protein structures, exceeding previous prediction techniques, the team says.

The bigger picture: DeepMind isn’t the only one working to accelerate scientific discovery with machine learning. Many other companies and researchers have sought to develop algorithms for discovering new drugs and new materials.