Table 1. A comparison of DendroBLAST with other tree inference methods on simulated multiple sequence alignments.
| A) Method of tree inference | RFd | FP | FN | Precision | Recall | |||||
| PhyML 4G SPR (SH 50) (3) | 46.2 | a | 12.2 | a | 34.0 | a | 0.943 | a | 0.863 | a |
| RAxML (3) | 46.8 | a | 23.4 | a | 23.4 | a | 0.906 | a | 0.906 | a |
| PhyML 4G SPR (3) | 49.4 | a | 24.7 | b | 24.7 | a | 0.900 | a | 0.900 | a |
| PhyML 4G (SH50) (3) | 57.7 | a | 17.5 | a | 40.2 | a | 0.918 | a | 0.838 | a |
| FastTree (SH50) (3) | 60.4 | a | 23.1 | a | 37.2 | a | 0.898 | a | 0.850 | a |
| PhyML 4G (3) | 63.4 | a | 31.7 | b | 31.7 | a | 0.872 | b | 0.872 | a |
| FastTree (3) | 64.6 | a | 32.2 | b | 32.3 | a | 0.870 | b | 0.870 | a |
| PhyML 1G (SH50) (3) | 69.1 | a | 24.7 | b | 44.4 | a | 0.888 | a | 0.821 | a |
| PhyML 1G (3) | 75.1 | a | 37.5 | c | 37.5 | a | 0.849 | b | 0.849 | a |
| DendroBLAST (1) | 95.1 | 27.6 | 67.5 | 0.867 | 0.732 | |||||
| FastME SPR (2) | 96.5 | b | 48.3 | c | 48.3 | a | 0.805 | c | 0.805 | a |
| FastME (2) | 101.2 | c | 50.6 | c | 50.6 | a | 0.796 | c | 0.796 | a |
| QuickTree log cor. (2) | 119.4 | c | 59.7 | c | 59.7 | a | 0.759 | c | 0.759 | a |
| QuickTree (con50) (2) | 142.6 | c | 67.5 | c | 75.0 | c | 0.717 | c | 0.697 | c |
| QuickTree (2) | 149.9 | c | 74.9 | c | 74.9 | c | 0.698 | c | 0.698 | c |
A) Comparison of DendroBLAST with several alignment based tree inference methods. B) Comparison of tree inference using FastME with distance matrices derived using different methods. In both cases (1) indicates that the tree inference method is distance-based and alignment-free, (2) indicates method is distance-based and alignment-dependent and (3) indicates alignment-dependent. RFd is the Robinson-Foulds distance. FP is the number of false positive bipartitions. FN is the number of false negative bipartitions. StdE is one standard error of the mean. Con 50 indicates that a majority-rule consensus tree was used to evaluate inference performance. SH 50 indicates that only bipartitions supported by an SH test scoring greater than 0.5 were used for evaluating the inference performance. 4G indicates that 4 gamma rate categories were used. SPR is subtree pruning regrafting. “a” indicates performance is significantly better than DendroBLAST, “b” indicates that performance is not significantly different to DendroBLAST and “c” indicates that performance is significantly worse than DendroBLAST. Performance evaluated by paired t-test at p≤0.001. Methods are ordered according to their Robinson-Foulds distance score.