Sign In

Communications of the ACM

ACM TechNews

Making New Drugs With a Dose of AI

View as: Print Mobile App Share:
DeepMind predicted the physical structure of a protein through the use of neural networks.

DeepMind, the artificial intelligence lab owned by Googles parent company, won the Critical Assessment of Structure Prediction contest.

Credit: T.M. Detwiler

An artificial intelligence (AI) designed by Google's DeepMind unit recently won the Critical Assessment of Structure Prediction competition, which concentrates on the mechanics of protein-folding to simplify drug discovery.

The victory stemmed from DeepMind researchers' design of a game to simulate protein-folding, followed by an AI that taught itself to play the game.

The deep learning algorithm predicted protein shapes by analyzing thousands of proteins, and it improved predictive accuracy by almost twice as much as experts anticipated from the winner of the contest.

Such breakthroughs rely on enormous volumes of computing power, with DeepMind benefiting from Google's underlying data centers, as well as the work of AI scientists in its employ.

Said DeepMind's Demis Hassabis, "It allows us to be much more creative, to try many more ideas, often in parallel."

From The New York Times
View Full Article - May Require Paid Subscription


Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA


No entries found