Figure 1.
Schematic illustrating the dependence of pathogen phylogenies on both the host contact network and the pathogen's population dynamics.
Route A: direct effect of the host contact network on the pathogen phylogeny. Route B: the host contact network changes the population dynamics, sometimes dramatically, and this in turn affects the pathogen phylogeny.
Figure 2.
Trees constructed from pathogens spreading on the NATSAL and ER networks supporting very different epidemic trajectories.
A: Phylogenetic tree from the NATSAL network, corresponding to the pathogen prevalence in panels D and E (blue lines). B and C: tree derived from the pathogen spreading on an ER network, corresponding to the red lines in D and E respectively. The ER prevalence was varied by changing the transmission parameter. Sampling was done at time 260 weeks.
Figure 3.
Mean number and mean sizes of clusters of two or more samples in phylogenetic trees from the scenarios.
A: cluster numbers (dynamic with duration 40 weeks); B cluster numbers (static network); C cluster sizes (dynamic with duration 40 weeks); D cluster sizes (static network).
Figure 4.
Comparison of typical trees derived from ER-like networks (top row) and NATSAL networks (middle row) illustrating that pathogen prevalence (bottom row) as well as networks both influence trees.
The NATSAL trees displays early divergence compared to the ER trees, and this affects the number of clusters. Panel A shows different epidemic trajectories and their corresponding trees, B shows more similar trajectories, and C shows closely matched epidemics. The tree differences are most modest in panel C where the pathogen population dynamics are closely matched. Edges in each cluster are drawn with the same colour. The threshold value for clustering was 0.1.
Figure 5.
Network prevalence, incidence and 1-cumulative degree distributions for ER (red) and NATSAL (blue) dynamic networks with .
Note that the NATSAL network admits similar epidemic trajectories with markedly different degree distributions (A–C). Panels D, E show the cumulative distributions of the cluster sizes in the ER (red) and NATSAL (blue) networks, and illustrate that these do not parallel the degree distribution; NATSAL networks do not have particularly more variable cluster sizes within trees. Panels F and G show the variance and skewness in boxplots; each box represents all trees from the given network and time point as in other figures.
Figure 6.
Branch lengths in phylogenetic trees from the scenarios.
A: mean branch lengths in trees from dynamic networks with a pathogen with duration of infectiousness 40 weeks; B mean branch lengths in static networks with duration of infectiousness 40 weeks. C, D: ratio of mean branch length to total tree distance, from dynamic and static networks.
Figure 7.
Mean internal/external branch lengths for trees derived from epidemics on dynamic (d = 40 weeks) and static networks.
Figure 8.
Tree imbalance in phylogenetic trees from the scenarios (left) dynamic with duration 40 weeks; (right) static networks.
Dashed lines indicate the expected imbalance for trees of this size [51].
Figure 9.
Cluster count, branch length and imbalance (top row) for a pathogen with duration of infectiousness d = 40, taken from simulations in which incidence and prevalence were as closely matched as possible.
Dashed line indicates the expected imbalance for trees of this size [51]. Prevalence and incidence over time in an ER network (blue) and NATSAL-based network (red) are shown in the bottom row for dynamic ER and NATSAL underlying contact networks. The number of lineages through time (LTT) in the trees for ER (solid) and NATSAL (dotted) is also shown. The LTT plots show the LTT for all trees; mean LTT at each time are indicated with dotted and solid lines and the coloured regions range from the minimum to the maximum. Distributions were close to uniform over this range. The ranges almost entirely overlap.
Figure 10.
Mean leaf-to-leaf distance scaled to the total distance in each tree, for the matched prevalence scenario, unmatched dynamic (
) and static.
Figure 11.
Clustering, branch lengths and imbalance for heterochronous sampling on the dynamic network with .
Clustering was done with a cut-off distance of 0.06 as in the results for homochronous sampling. The expected value of imbalance is 0.074 [51], considerably less than the imbalance of the heterochronously sampled trees from both networks.