Skip to main content


Figure 4 | BMC Evolutionary Biology

Figure 4

From: An empirical evaluation of two-stage species tree inference strategies using a multilocus dataset from North American pines

Figure 4

Principal components analysis (PCA) and multidimensional scaling (MDS) of phylogenetic inference strategies. PCA was applied to 72 phylogenetic inference strategies, in which each strategy is viewed as a point in a 145-dimensional space of clades. MDS was applied to a distance matrix between all pairs of strategies, where the distance between a pair of strategies was the mean Robinson-Foulds distance between all pairs of bootstrap trees from one strategy and another strategy (i.e., mean of 106 comparisons). The plots show the first and second components. On a given row, each plot represents the same 72 points in the space of components 1 and 2; the three plots are colored differently to highlight different features. (A) Colors represent different methods for constructing species trees. (B) Colors represent different gene tree inference methods. (C) Colors represent different outgroups. The points on each graph represent different combinations of the three factors that form phylogenetic inference strategies. Each line in part A represents the resultant vector (scaled by a constant to lie within the span of the 72 points) for all 12 points of a certain method for constructing species trees. Each line in part B represents the resultant vector for all 18 points of a gene tree inference method (scaled by a constant). Each line in part C represents the resultant vector for all 24 points of an outgroup (scaled by a constant). The scaling constants in parts A, B, and C are distinct. Each of the shaded regions in parts A and B is a convex hull of the points from a particular method.

Back to article page