Has DeepMind’s AI Solved One of Biology’s Biggest Challenges?

Keegan Schroeder
8th December 2020

One of DeepMind’s AI, AlphaFold, has been making waves in the field of Structural Biology after predicting multiple protein structures with accuracy comparable with currently used lab techniques. This news follows AlphaFold attending and winning 2020s Critical Assessment of Techniques for Protein Structure Prediction (CASP), which can almost be considered the “World Cup” for communities researching protein structure prediction methods.

Each living cell contains a staggering number of different proteins. The scientific community has recognised over 200 million proteins, but the structures are only known for a small fraction. These are essential for almost every function the cell carries out. As such, abnormalities in proteins form the foundation for almost every human disease. Proteins have a very specific shape that dictates their function. This comes about as a result of folding (think molecular origami).  For example, a protein called TP53 is altered in over half of all human cancers and is a major focus for research.

Traditional techniques for figuring out protein structure involve extremely time-consuming work from structural biologists in a lab, deploying techniques such as NMR (Nuclear Magnetic Resonance) or Mass Spectroscopy. The AI’s ability to predict protein shape marks a tremendous leap forward in molecular biology.

The door to new drug products from the field of drug design is beginning to close as researchers struggle to keep pace with demand. There is hope that the huge leap taken by AlphaFold will mark the opening of a new door onto the world of protein design, paving the way to “green enzymes” that can break down plastic waste, and a huge range of new medicines.