This is an updated version.
Head Image Caption: SARS-CoV-2 Membrane Protein: rendering of one of DeepMind’s protein structure predictions
In a bid to help the global research community better understand the coronavirus, DeepMind today released the structure predictions for six proteins associated with SARS-CoV-2, the virus that causes COVID-19, using the most up-to-date version of their AlphaFold system.
As the world struggles with the COVID-19 outbreak, one research team after another in the global scientific community has stepped up to offer expertise, tools and possible solutions. In the early stages of the outbreak front-line labs open-sourced genomes of the virus which enabled other researchers to rapidly develop tests around the pathogen. Other labs modelled the coronavirus infection peak or produced molecular structures to develop drug compounds and treatments against infection.
Understanding a protein’s structure provides an important resource for making sense of how a virus functions, but experiments to determine structure typically take months or longer. To speed things up, researchers have been developing computational methods to predict protein structure from amino acid sequences.
In January DeepMind published AlphaFold, a deep learning system that aims to accurately predict protein structure even when no structures of similar proteins are available and generates 3D models of proteins with SOTA accuracy. DeepMind says it has continued to improve AlphaFold for better predictions since the release, and they’re now able to share with the public the predicted structures for some of the proteins in SARS-CoV-2 generated with their newly-developed methods.
“We confirmed that our system provided an accurate prediction for the experimentally determined SARS-CoV-2 spike protein structure shared in the Protein Data Bank,” DeepMind researchers wrote in their official blog. “We recently shared our results with several colleagues at the Francis Crick Institute in the UK, including structural biologists and virologists, who encouraged us to release our structures to the general scientific community now.”
Although the current structure predictions have not yet been peer-reviewed or experimentally verified, given the seriousness and time-sensitivity of the situation, DeepMind decided to release the predicted structures now in the hope the work can contribute to the scientific community’s interrogation of how the virus functions and provide a hypothesis generation platform for future experimental work in developing treatments.
The structure predictions, relevant technical details and data can be downloaded here.
DeepMind continues to improve the AlphaFold system, and the UK-based AI company and research lab today announced the release of an updated version of its SARS-CoV-2 structure predictions for five understudied SARS-CoV-2 targets: SARS-CoV-2 membrane protein, Nsp2, Nsp4, Nsp6, and Papain-like proteinase.
On June 17, UC Berkeley’s Brohawn lab released an experimental structure of the SARS-CoV-2 ORF3a protein (Protein 3a) in the Protein Data Bank (PDB). As this experimentally obtained structure and the structures predicted by the AlphaFold system earlier this year were in very good agreement, DeepMind was inspired to release its new set of predictions for the remaining five proteins that have not yet been experimentally determined.
The most up-to-date structure predictions can be downloaded here.
Journalist: Yuan Yuan; Fangyu Cai | Editor: Michael Sarazen