DeepMind Publishes AlphaFold, an AI Focused on Solving the Protein Folding Problem

Designed to solve a key problem in biology, AlphaFold has entered the charts as the most accurate protein folding predictor yet.

DeepMind's AlphaFold has proven accurate in predicting protein folding. (📷: Senior et al)

Google's DeepMind division has published a study on using artificial intelligence to drive scientific discovery, turning its AlphaFold AI to the problem of predicting protein structure from large datasets.

"In our study published today in Nature, we demonstrate how artificial intelligence research can drive and accelerate new scientific discoveries," explain authors Andrew Senior, John Jumper, and Demis Hassabis. "We've built a dedicated, interdisciplinary team in hopes of using AI to push basic research forward: bringing together experts from the fields of structural biology, physics, and machine learning to apply cutting-edge techniques to predict the 3D structure of a protein based solely on its genetic sequence.

"Our system, AlphaFold – described in peer-reviewed papers now published in Nature and PROTEINS – is the culmination of several years of work, and builds on decades of prior research using large genomic datasets to predict protein structure. The 3D models of proteins that AlphaFold generates are far more accurate than any that have come before—marking significant progress on one of the core challenges in biology."

While the folding of protein chains is a key aspect of biology, it is difficult to predict: Previous attempts, including the popular distributed computing effort Folding@Home, have largely centred around throwing increasing computational power at the problem. AlphaFold is different: It uses neural networks to predict physical properties, and does so with considerable accuracy.

"The success of our first foray into protein folding is indicative of how machine learning systems can integrate diverse sources of information to help scientists come up with creative solutions to complex problems at speed," the team claims. "Just as we've seen how AI can help people master complex games through systems like AlphaGo and AlphaZero, we similarly hope that one day, AI breakthroughs will help serve as a platform to advance our understanding of fundamental scientific problems, too."

The paper, written in partnership with co-authors throughout DeepMind, the Francis Crick Institute, and University College London, is available through the journal Nature; AlphaFold's source code, written in Python and available under the permissive Apache 2.0 Licence along with Creative Commons Attribution-NonCommercial 4.0 International-licensed weights and data, can be found on the DeepMind GitHub repository.

Gareth Halfacree
Freelance journalist, technical author, hacker, tinkerer, erstwhile sysadmin. For hire:
Related articles
Sponsored articles
Related articles