Water buffalo, Sunda pangolins and mink are among the 540 mammals predicted to be likely to spread the coronavirus based on their biology and where they live
17 November 2021
The SARS-CoV-2 coronavirus, which causes covid-19, invades human and animal cells by engaging the ACE2 protein on host cells with its spike protein. This step is required to infect an animal and be transmitted to other hosts.
Distinct species have different versions of the protein, so understanding how well their ACE2 protein binds to the coronavirus spike protein can help scientists predict which animals are most likely to spread covid-19. But the amino acid sequences that make up the ACE2 protein are available for only around 300 species.
Barbara Han at the Cary Institute of Ecosystem Studies in New York and her colleagues built a machine learning tool to predict whether the ACE2 protein from 5400 mammalian species can bind strongly enough to the spike protein from the original coronavirus variant to spread the virus, even without knowing their ACE2 amino acid sequences.
The species predicted to be able to harbour the virus include white-tailed deer, which were recently found to have very high rates of infection in North America.
Striped skunk and 76 rodent species including rats and deer mice were also deemed likely to spread the coronavirus, along with some farmed species such as water buffalo.
To create the model, the team first estimated how strongly the spike protein binds to the ACE2 protein from 142 mammalian species for which the ACE2 sequences are known, and whether or not these species are likely to spread the coronavirus based on this binding strength.
They then trained the AI to learn patterns between transmissibility and a set of around 60 ecological and biological traits gathered from previous studies. The traits included where the animals live, how much their habitats overlap with human populations, their lifespan, how varied their diet is and their body mass.
When given biological and ecological trait data for the other species, the model could then guess the likelihood of different species being able to spread the coronavirus.
These results must be followed up with systematic surveillance and lab studies to test and validate the predictions, says Han.
“This is an incredibly useful approach to prioritise animal species for surveillance,” says Arinjay Banerjee at the University of Saskatchewan in Canada. Surveillance will help track viral infections and the possible emergence of animal-adapted coronavirus variants, says Banerjee.
Journal reference: Proceedings of the Royal Society B, DOI: 10.1098/rspb.2021.1651
More on these topics: