In a bid to speed up efforts to develop a vaccine, DeepMind have released structure predictions for several under-studied proteins associated with SARS-CoV-2, the virus that causes COVID-19, which have been generated using the latest version of their deep learning system, AlphaFold. This was achieved through training neural networks to predict the shape of a protein from its genetic sequence. These neural networks are optimized through ‘learning’, where small incremental improvements would be made with each simulation. DeepMind stress that that while their structure prediction system is still in development and they cannot be certain of the accuracy of the structures it has provided at this stage, they hope they may add to researchers’ understanding of SARS-CoV-2. Much of the protein folding has relied on
the genomic blueprint released by China early in the pandemic. Folding@Home makes their work open to the scientific community, whilst DeepMind’s release has been posted with open access, continuing a successful and ongoing culture of collaboration within the science community