Neural network predicted COVID-19 spreaders among animals

The neural network has identified 540 mammals that are most likely to be COVID-19 distributors. According to the model, minks, buffaloes and bats are among the 10% most likely distributors, which corresponds to the results of laboratory experiments.

The SARS-CoV-2 coronavirus, which causes COVID-19, penetrates human and animal tissues by interacting its spike protein with the ACE2 protein in the carrier cells. Different biological species have different versions of the ACE2 protein, so understanding how well their protein binds to the coronavirus protein can help predict which animals are most likely to become infected and possibly spread COVID-19.

However, the amino acid sequences that makeup ACE2 are known only for about 300 species. To circumvent this limitation, scientists from the Carey Institute for Ecosystem Research in New York have created a model predicting whether the ACE2 protein from 5,400 mammalian species can bind strongly enough to the coronavirus spike protein, even without knowing their ACE2 amino acid sequences.

Also, about 60 ecological and biological features of each species were included in the training data of the model. These features include the geography of the habitat, the degree of coincidence of the habitat with the human population, life expectancy, as well as the diversity of the diet and body weight of representatives of the species.

The resulting model estimated the probability of the spread of coronavirus by each of the 6495 mammalian species examined. The 10% of animals most likely included bats, lizards, minks, buffaloes, deer and 76 species of rodents. For some of these animals, there were already experimental confirmations of the possibility of spreading the virus, which indirectly confirms the correctness of the model estimates.

Notify of

Inline Feedbacks
View all comments