DeepMind Releases AI-Generated Protein Structures to Aid COVID-19 Research
DeepMind uses its advanced AlphaFold system to predict the structures of under-studied SARS-CoV-2 proteins. These computationally generated models offer a hypothesis-generating resource for scientists developing therapeutics.
DeepMind applies its latest AlphaFold deep learning system to predict the structures of several under-studied proteins associated with SARS-CoV-2, the virus that causes COVID-19. By releasing these computational models, the company aims to support the global scientific community's efforts to understand how the virus functions. This initiative builds upon the rapid open-access sharing of viral genomes and epidemiological data by researchers worldwide.
Traditional methods for determining protein structures through experiments often take months or prove entirely intractable, creating a bottleneck in the fight against the pandemic. AlphaFold addresses this challenge by using advanced algorithms to predict protein structures directly from amino acid sequences. The system specifically focuses on "free modelling," which allows it to make accurate predictions even when no similar protein structures currently exist in scientific databases.
The researchers emphasize that these structure predictions remain unverified by physical experiments and carry a degree of uncertainty. Despite this limitation, DeepMind expresses confidence that the system provides highly useful models that surpass previous computational methods. The team hopes these AI-generated structures serve as a valuable hypothesis generation platform to guide future experimental work in developing effective COVID-19 therapeutics.