Thread by @bahanbug

Very excited to share our latest work predicting mammal species with the capacity to transmit SARS-CoV-2: bit.ly/3qASgWs #covid19 @fischhoff @aacastell39 @jpglmrodrigues @Varsani_lab (1/10)

Background: Many mammals appear possible hosts for SARS-CoV-2 - new hosts confirmed each wk - but not all are equally susceptible. Some are susceptible but don’t shed virus, others demonstrate onward transmission to other animals, including humans. (2/10)

New animal hosts pose risk of secondary spillover, where SARS-CoV-2 moves from people to animals and back to people, as has already happened with farmed mink (bit.ly/3qFVLe0) (3/10)

We use the term 'zoonotic capacity' to describe the degree to which an animal can transmit SARS-CoV-2 onward to other animals or people. This is a requirement for SARS-CoV-2 to persist in animal hosts and seed new spillovers to humans. (4/10)

Predicting species with high zoonotic capacity has been fundamentally constrained by missing data (ACE2 sequences) and the impossibility of running SARS-CoV-2 infection experiments on every mammal. (5/10)

We expand this very limited predictive capacity by more than a factor of 10 (to 5,400 species), by using #MachineLearning to combine data on mammal traits (like body size or lifespan) with 3D models of how the virus and host cell proteins interact. #multiscale (6/10)

Our model predicted zoonotic capacity with 72% accuracy, and matched up with IRL results for white-tailed deer, mink, raccoon dogs, and others; and identified many species likely to have high zoonotic capacity. (7/10)

Notably, many predicted species are in close contact with people (captive, farmed, pets, live traded, hunted), and survive well in human-altered landscapes (including living around people’s homes). (8/10)

Close contact with people is important because SARS-CoV-2 infection in animals will primarily come from humans (spillback transmission). We mapped where this is most likely to happen (in human COVID-19 hotspots). (9/10)

There are many efficiencies to be gained by coordinated iteration between field surveillance, lab work, & computational predictions like these. Synergies across disciplines & methods will be paramount for pandemic response now and in the future. (10/10)