AI unravels one of the great challenges of biology

For decades, scientists have been trying to figure out how to quickly predict the twisted and tangled shape of proteins and, from there, unravel a greater understanding of the machinery of life itself.

This week, an artificial intelligence program created by Google’s sister company DeepMind was shown to practically solve the challenge, predicting the way proteins contort into three-dimensional structures in the results of a biannual competition that the judges hailed as a game changer.

“In a sense, the problem is solved,” said John Moult, a computational biologist at the University of Maryland, who co-founded the Critical Assessment of Techniques for Protein Structure Prediction (CASP) competition in 1994, as quoted by the journal. Nature.

Janet Thornton of the European Bioinformatics Institute said in a statement from the competition organizers on Monday that the work to solve the problem was a “triumph of human curiosity, effort and intelligence.”

“A better understanding of protein structures and the ability to predict them using a computer means a better understanding of life, evolution, and of course human health and disease,” said Thornton, who is not affiliated with CASP. or DeepMind.

The cells of all living things contain thousands of proteins, workhorses that catalyze most of the body’s chemical reactions.

They are essential to life, from muscle function to the transport of oxygen in the blood, and they are also the key to diseases like cancer and even COVID-19.

A protein begins as a chain of amino acids, which is then crumpled into a unique three-dimensional tangle.

It is this form that is directly related to its function.

Scientists have been puzzled for half a century about how to accurately and quickly predict what formation (among an unfathomable number of possibilities) a protein might take by looking at its chain of amino acids, a process that can take years in the laboratory.

The CASP competition involved around 100 teams who were given the amino acid sequences of dozens of proteins and tasked with estimating their final shapes, which were known to the organizers.

DeepMind, whose AlphaGo program stunned the world in 2016 by beating an elite human gamer in the complex strategy game “Go,” was already on top of the field in the last contest of 2018.

This time, his AlphaFold program determined the shape of many of the proteins “to a level of precision comparable to that achieved with expensive and time-consuming laboratory experiments,” according to the CASP organizers.

‘This changes medicine’

Andrei Lupas, an evolutionary biologist at the Max Planck Institute for Developmental Biology who was part of the evaluation team, told Nature that AlphaFold had helped him determine the structure of a protein that his lab had been trying to pin down for a decade.

“This will change medicine. It will change research. It will change bioengineering. It will change everything,” he told Nature.

Derek Lowe, writing on drug discovery and the pharmaceutical industry for Science Translational Medicine, described protein folding as “watching hinged piles of lumber spontaneously pile up into functional boats, wagons and treehouses.”

He said the AlphaFold results did not mean that the program would consistently generate the correct protein structure.

“But getting that level of structural precision in so many varied proteins is something that has never been done before.”

DeepMind said it was looking at how the program could help increase awareness of certain diseases, for example to identify whether a protein has malfunctioned.

“These insights could allow for more precise work in drug development, complementing existing experimental methods to find promising treatments faster,” he said in a statement.

The firm added that it was working to produce a peer-reviewed article and was “exploring how best to provide wider access to the system.”

© 2020 AFP