Fig. 8.

Patterns of CorrTest accuracy using the specific thresholds determined by predictive models trained with different ranges of the number of sequences (n): M100 (n ≤ 100), M200 (100 < n ≤ 200), M300 (200 < n ≤ 300), and M400 (n > 300) for the corresponding test data sets (Tamura et al. 2012). Accuracies are shown for 50, 100, 200, 300, and 400 ingroup sequences. The accuracy of CorrTest for different sequence lengths is shown when (a) the correct topology was assumed and (b) the topology was inferred. (c) The accuracy of CorrTest for data sets in which the inferred topology contained small and large number of topological errors. Darker color indicates higher accuracy.