Phylogeographic and phylodynamic analyses
To
infer significant transition rates between epidemiologic features (known
as ‘traits’) of PB2 sequences
along phylogenetic tree branches,
including (a) geographic location, (b) host species, and (c) HA and
Neuraminidase (NA) subtypes, and evaluate continuous virus diffusion
(rate of movement in km/year) among host taxa, joint discrete and
continuous trait Bayesian phylogeographic and phylodynamic analyses were
performed using BEAST v.1.10.4 (Suchard et al., 2018). An asymmetric
substitution model with Bayesian stochastic search variable selection
(BSSVS) and a strict clock model were used to estimate diffusion between
discrete states (Lemey, Rambaut, Drummond, & Suchard, 2009).
Statistical significance for diffusion between discrete states was
determined by the following Bayes
Factors (BF): 3 ≤ BF < 20, 20 ≤ BF <150, and BF ≥
150, denoting positive, strong, and very strong support, respectively
(Kass & Raftery, 1995). To quantify transitions among geographic, host,
or HA/NA subtype states (Markov jumps, i.e. the frequency of transitions
from one geographic or host state to another along phylogenetic
branches) and the duration of time viruses spend in each state (Markov
rewards), posterior inference of the complete Markov jump history
through time was included (Hong, Lemey, Suchard, & Baele, 2021).
Continuous Bayesian phylogeographic analyses were performed by
annotating the empirical tree with latitude and longitude locations of
all sampling sites via www.geonames.org to infer and visualize the
evolutionary and geographic relationships between global sequences.
Sequences between 1978 and 2008 were masked so that they contributed to
the overall tree structure but were not included in the calculation of
diffusion rates globally or within Iceland (Hicks, Dimitrov, Afonso,
Ramey, & Bahl, 2019). A Cauchy relaxed random walk (RRW) model and an
uncorrelated relaxed lognormal clock were used and a random jitter of
0.001 was applied to isolates with matching geographic latitude and
longitude coordinates (Lemey et al., 2010). Host-specific diffusion
rates, the rate of virus movement (in kilometers per year), were
calculated (Dellicour, Rose, & Pybus, 2016; Minin & Suchard, 2008;
Trovão, Suchard, Baele, Gilbert, & Lemey, 2015). GIS mapping was
performed using QGIS v.3.16.3 (QGIS Development Team, 2020).