In the following, we present two published examples of virus phylogeographic studies re-examined with PastView. The first one is related to the HIV-1A epidemiological history in Albania. The second is related to the global phylogeography of Dengue serotype 1 virus.
Example 1
Study of HIV-1A epidemiological history in Albania.
Figure 4 shows a simple example of PastView output, in the study of HIV-1A epidemiological history in Albania. The dataset, tree, and origin of the sequences (locations) are from [15] (see also [13]). Ancestral locations are computed by parsimony (DELTRAN), maximum of the marginal posterior probabilities (MAP, F-81 like), and joint most likely scenario. Computation time of ancestral annotations for this small dataset (153 strains) is ~ 2 s with a 3.1 GHz Intel I7 computer. The tree is foreground color-coded if the ancestral annotations are identical between MAP F-81 like, joint most likely reconstruction and parsimony. If not, color-coded bubbles are displayed (Fig. 4a) (in the same order, reading from root to tips). Here, we identify nine nodes with discrepancies, and only four nodes if we compare MAP and joint inferences only. The pie charts display (Fig. 4b) the posteriors for nodes having two or more annotations with probability higher than their MAP F81 like probability value minus 40%. Based on MAP and joint inferences, the transition maps (Fig. 4c and d) are slightly different but output the same global scenario: the virus spread from Africa to West Europe, East Europe and Greece, and then from Greece to Albania, with a few Greek sequences coming back from Albania. This reaches the same general conclusion as [15]: there has been a single major introduction of HIV-1A from Greece followed by a local epidemic spread. This result (Fig. 4e) is highlighted by the thick pathways from the tree root to all entries to Albania.
Example 2
Study of the global phylogeography of dengue serotype 1 virus.
The second example is related to the study of the worldwide phylogeography of Dengue virus of serotype 1 (DENV-1). The dataset, tree (269 strains), and origin of the sequences (13 locations) were obtained from [16]. Tree and ancestral annotations are first computed with BEAST (same parameters / model as [16]), then imported in PastView. Another set of ancestral annotations is computed with PastView (F81-like method, MAP, computation time is ~40s with an 3.1 GHz Intel I7 computer). The tree is then foreground color-coded (Fig. 5a) if the ancestral annotations (MAPs) are the same between the two methods. If not, pie-charts are displayed (from a root to tips reading: left = BEAST, right = PastView). Both analyses reach the same main conclusions as described in [16]: Southeast Asia countries are found to be the most likely origin of the virus dispersion and India played a crucial role in the establishment, evolution and dispersal of the Cosmopolitan (Africa, America, Carribean, East & Southeast Asia) DENV-1 genotype (Fig. 5b and c). The Caribbean region is also found by both methods as the dissemination origin point of the virus to the Americas (South and Central America, then North America). If there is a global consensus for the most ancestral nodes/ transitions, some differences between the results obtained with the two methods exist especially for the most recent nodes/transitions (Fig. 5a, b and c). These disagreements are related mainly to transitions within small clusters of sequences with heterogeneous annotations.