3D structures of almost 200 million proteins have finally been predicted | Science and technology news
Artificial intelligence company DeepMind has announced a major medical-scientific breakthrough in determining the structures of nearly 200 million proteins.
Proteins are not two-dimensional molecules, but instead have chemical properties determined by their three-dimensional shape — but figuring out those shapes is an intense process.
The breakthrough has significant implications for medicine, and the new research from Google-backed DeepMind is being hailed as having “the potential to dramatically advance our understanding of biology.”
A protein is made up of a chain of amino acids, but without knowing how these chains are connected it is not possible to know how they can interact with human cells and be modified.
Over the past year, DeepMind, owned by Google parent company Alphabet, shared the fruits of an AI system called AlphaFold which could predict the 3D structure of a protein from its one-dimensional amino acid sequence.
A year earlier, PC gamers had been asked to do so donate part of their computing power to an international effort to study diseases like COVID-19 and Alzheimer’s to simulate the molecular dynamics of protein folding.
It is such an important topic for medical science because the structure of proteins determines chemical reactions in human cells and thus the entire human body – but until now only a fraction of the protein structures were known.
DeepMind’s announcement and protein structure database, which is open to the public, dramatically increases the number of known protein structures from nearly one million to over 200 million.
It was developed together with EMBL’s European Bioinformatics Institute (EMBL-EBI), whose Director General Edith Heard said: “AlphaFold now offers a 3D view of the protein universe.”
“We have been amazed at the speed at which AlphaFold has already become an indispensable tool for hundreds of thousands of scientists in laboratories and universities around the world,” said Demis Hassabis, founder and CEO of DeepMind.
“From fighting disease to tackling plastic pollution, AlphaFold has already enabled an incredible impact on some of our biggest global challenges,” added Mr. Hassabis.
“We hope that this expanded database will support countless more scientists in their important work and open completely new avenues of scientific discovery.”
The research has been welcomed by scientists who have used AlphaFold models to develop anti-malarial antibodies and even special enzymes that could break down plastics.
Since its launch, more than 1,000 scientific papers have been published and the database has been accessed by over 500,000 researchers from over 190 countries.
Other research areas made possible by the database include honey bee health, understanding ice formation and neglected diseases such as Chugs disease and leishmaniasis.
“This is just the effect of a million predictions; imagine the impact when over 200 million protein structure predictions are openly accessible in the AlphaFold database,” said Sameer Velankar, who leads the protein database team at EMBL-EBI in Europe.