We likely can't do this until we have the full prevalence data, but can we get descriptors going?
Methods we're likely to try, and who can implement them
- Han reservoir approach with BRTs (Becker, Han?, Dallas?)
- Plug and play (Dallas)
- BART from sequence-munging + host traits (Carlson, Albery? Who wants to do sequences with me?)
- Elmasri latent trait approach (Farrell)
- Linear filtering (Poisot)
- k-NN (Poisot)
Goals
- adapt the Eng methods from influenza A
- try to classify zoonotic risk among coronaviruses, and maybe flaviviruses
Needs
- who can help wrangle the actual sequence data and process it into features?
Coming up with a spatial host-virus network can be done either in a coarse way like the synthetic food web paper, or in a slightly more fancy way like in the Elton & Grinnell paper. Long story short, can we map some properties of the network over space? The ingredients for this are (i) species distributions and (ii) a network, possibly augmented by inferred links. I feel like this can go a long way towards a map of spillover risk?