Artificial intelligence tool cracks code to imagine proteins in 3D
An artificial intelligence network solved a scientific problem that has stumped researchers for half a century, by successfully predicting the way proteins fold into three-dimensional shapes, a process that has generally required painstaking and expensive laboratory work that it could go on for decades.
The way proteins, one of the building blocks of life, fold up drives their functionality and behavior. For example, SARS-Cov-2 has a protein that folds like a spike. This form, therefore, is relevant to biologists (even for its ability to find cures for diseases). However, it is not easy to predict the shape of a protein based on how amino acids join together to form a protein. That’s because there are countless ways a protein can fold into a three-dimensional structure.
Google-owned DeepMind created a computer program called AlphaFold, which predicted with astonishing precision the three-dimensional shapes of proteins after being fed their constituent parts – data representing chains of amino acids.
“This computational work represents an amazing advance in the problem of protein folding, a great 50-year challenge in biology. It has happened decades before many people in the field would have predicted it. It will be exciting to see the many ways in which biological research will fundamentally change, ”said Professor Venki Ramakrishnan, Nobel Laureate and President of the Royal Society, according to a DeepMind blog post.
“It is a breakthrough of the first order, without a doubt one of the most important scientific results of my life,” said a Nature report, quoting Mohammed Al Quraishi, a computational biologist at Columbia University in New York City. “I think it’s fair to say that this will be very disruptive to the field of protein structure prediction. I suspect that many will leave the field, since the central problem could be said to have been solved, ”he added.
Quraishi was part of the Critical Assessment of Structure Prediction (CASP), a competition held every two years to accelerate research in the field, where AlphaFold reached the threshold of what is considered to have “solved” the problem.
DeepMind became a subsidiary of Google after an acquisition in 2014 and is best known for its AI for Gamers, which was taught to beat Atari video games and world-renowned Go players like Lee Sedol. The company’s ambition has been to develop AI that can be applied to broader problems, and so far it has created systems to make Google’s data centers more energy efficient, identify eye disease from scans, and generate speech with human sound.
“These algorithms are now becoming strong and powerful enough to be applicable to scientific problems,” said DeepMind CEO Demis Hassabis in a call with reporters, the Bloomberg news agency reported. After four years of development, “we have a system that is accurate enough to really have biological significance and relevance to biological researchers.”
The DeepMind blog post referenced comments from eminent scientists on the subject in the past to illustrate the importance of advancement. “In his 1972 Nobel Prize in Chemistry acceptance speech, Christian Anfinsen famously posited that, in theory, the amino acid sequence of a protein should completely determine its structure. This hypothesis sparked a five-decade quest to be able to computationally predict the 3D structure of a protein based solely on its 1D amino acid sequence as a complementary alternative to these expensive and time-consuming experimental methods, “he said.
However, a major challenge is that the number of ways a protein could theoretically fold before settling into its final 3D structure is astronomical. In 1969, Cyrus Levinthal pointed out that it would take longer than the age of the known universe to list all the possible configurations of a typical protein by calculating brute force. Levinthal estimated 10 ^ 300 possible conformations for a typical protein, “he added.
CASP scientists analyzed the shape of the amino acid sequences of about 100 proteins. Competitors received the sequences and were in charge of predicting their shape.
AlphaFold’s evaluation aligned almost perfectly with the CASP analysis of two-thirds of the proteins, compared to about 10% of the other teams, and better than what the DeepMind tool achieved two years ago.
Hassabis said his inspiration for AlphaFold came from “citizen science” attempts to find unknown protein structures, such as Foldit, which presented the problem to amateur volunteers in puzzle form.
In their first two years, human players proved surprisingly good at solving puzzles and ended up discovering a structure that had puzzled scientists and designing a new enzyme that was later confirmed in the lab.