A group of scientists from the Univerity of Liverpool have enabled machine learning (ML) technology to understand and prevent new forms of coronaviruses from spreading by understanding which animals could be the next hosts for the virus's new forms. 

They have used ML to identify the animals through three different yet complementary perspectives - viral, mammal and network. This bifurcation allowed the scientists to segregate high-risk species to keep a lookout for coronavirus spread. 

Of what we know of the coronaviruses, they're a family of viruses that are capable of affecting both - birds and mammals. Therefore, the researchers used ML to triangulate data of GenBank’s 411 strains of coronavirus, the National Institutes of Health database, and 876 host species of mammals and tried to predict a relationship between them. The calculations identified the animals that are most likely to be infected by the coronaviruses, which could lead to further mutations of the virus. The ML calculations pointed out that there are 11.5 times more associations between species of mammals and strains of coronavirus than previously known.

The ML identified 126 species that can be hosts to the virus - the Asian civet is predicted to be a host to 32 coronaviruses; the horseshoe bat to 67; the intermediate horseshoe bat to 44; and the pangolin to 14. However, the absolute champion of SARS-CoV-2 recombination hosting was the domestic pig, capable of harbouring 121 new types of coronavirus.