Transmission and phylogenetic trees
We constructed transmission trees between pumas in each region using the
R package TransPhylo
(Didelot et
al. 2017). TransPhylo uses a time-stamped phylogeny to estimate a
transmission tree to gain inference into “who infected whom” and when.
Briefly, this approach computes the probability of an observed
transmission tree given a phylogeny using a stochastic branching process
epidemiological model; the space of possible transmission trees is
sampled using reversible jump Markov chain Monte Carlo (MCMC)
(Didelot et
al. 2017). This approach is particularly useful for pathogens where the
outbreak is ongoing, and not all cases are sampled
(Didelot et
al. 2017), as is the case here. We leveraged our FIVpcoBayesian phylogenetic reconstructions from previous work and focused on
the two clades of FIVpco that predominantly occurred in
each region (see Fountain-Jones et al. 2019). Whilst the
TransPhylo approach makes few assumptions, a generation time
distribution (the time from primary infection to onward transmission) is
required to calibrate the epidemiological model
(Didelot et
al. 2017). We assumed that generation time could be drawn from a Gamma
distribution (k = 2, θ = 1.5) estimating onward transmission on average
3 years post-infection (95% interval: 0.3 - 8 years, based on average
puma age estimates (Logan & Sweanor 2001)). Based on previous work
(Lewis et
al. 2015), we were confident that the proportion of cases (π) sampled
was high, therefore we set the starting estimate of π to be 0.6 (60% of
cases tested in each region), and allowed it to be estimated by the
model. We ran multiple MCMC analyses of 400,000 iterations and assessed
convergence by checking that the parameter effective sample size (ESS)
was > 200. We computed the posterior distributions of
R0, π, and the realized generation time from the MCMC
output. We also estimated likely infection time distributions for each
individual and compared these estimates to approximate puma birth dates
to ensure that these infection time distributions were biologically
plausible. We then computed a consensus transmission tree for each
region to visualize the transmission probabilities between individuals
through time. Lastly, we reformatted the tree into a network object
(nodes as individual puma and edges representing transmission
probabilities) and plotted it using the igraph package (Csárdi &
Nepusz 2006) and overlaid puma sex as a trait. Overall weighted degree
and weighted degree for each sex, including edges representing homophily
(e.g., male-male) and heterophily (e.g., male-female), were also
calculated using igraph .