Estimating host breadth and cross-species transmission by parasite
species
We considered three network-level factors for each parasite species: (1)
the number of different host species a parasite infected in a community
(host breadth), (2) the number of predicted transmission pairs between
different host species (cross-species transmission connections), and (3)
a combination of the two (graph density). Calculating each of these
required that we first construct networks of the host species infected
by a particular parasite species (examples in Fig. 2). To do this, for
each parasite species, we constructed undirected networks (i.e.,
edges/connections do not have a direction) of host species in each site
where each node represented the population of a host species in a given
community. With only the host species that were present in that site as
the nodes, we added edges between host species if they had overlapping
epidemics of that same parasite species. For the networks used to
calculate host breadth, we added node edges to itself (loops) for each
host that was infected by a given parasite (these loops are equivalent
to filling a diagonal cell in the adjacency matrix). We then calculated
host breadth as the proportion of all hosts in a site (observed across
all time points) that were infected with a given parasite species. The
networks used to estimate cross-species transmission did not contain
loops (that is, node edges from a host species back to itself), because
we were only interested in interconnections between species. We then
estimated cross-species transmission using the proportion of realized
connections between host species out of all possible connections in the
network. Finally, because both of these metrics are important for
different reasons, we combined the two and calculated graph density as
the proportion of realized edges out of all potential edges that could
occur in a given network, including loops where a host connects to
itself if it was infected. Additionally, since sites varied in their
host community composition, we only included a host species in a given
network if that species was found in that site; sites where the parasite
was not found were dropped from the analysis.