Table 4 Discrimination capacities of the downsampled bootstrap support values

	Bootstrap combinations and resulting average AUC values
Ref. branches	IMD	ME	ME + IMD	ML	ML + IMD	ME + ML	ME + ML + IMD	Number of datasets
IMD80	0.940 ± 0.06	0.783 ± 0.16	0.923 ± 0.07	0.794 ± 0.16	0.923 ± 0.08	0.796 ± 0.15	0.906 ± 0.09	55
ME80	0.810 ± 0.16	0.830 ± 0.14	0.860 ± 0.14	0.867 ± 0.11	0.873 ± 0.13	0.852 ± 0.13	0.869 ± 0.13	55
ME80 ∩ IMD80	0.910 ± 0.10	0.849 ± 0.12	0.922 ± 0.10	0.870 ± 0.13	0.926 ± 0.10	0.863 ± 0.13	0.915 ± 0.10	55
ML80	0.829 ± 0.15	0.794 ± 0.14	0.858 ± 0.14	0.843 ± 0.14	0.880 ± 0.13	0.832 ± 0.14	0.868 ± 0.13	54
ML80 ∩ IMD80	0.926 ± 0.09	0.822 ± 0.14	0.929 ± 0.07	0.848 ± 0.14	0.939 ± 0.07	0.841 ± 0.14	0.918 ± 0.08	52
ME80 ∩ ML80	0.833 ± 0.14	0.842 ± 0.12	0.880 ± 0.11	0.882 ± 0.10	0.902 ± 0.09	0.869 ± 0.11	0.892 ± 0.09	55
ME80 ∩ ML80 ∩ IMD80	0.911 ± 0.11	0.856 ± 0.13	0.929 ± 0.10	0.874 ± 0.13	0.934 ± 0.10	0.867 ± 0.14	0.923 ± 0.10	52
Average	0.880	0.825	0.900	0.854	0.911	0.846	0.899

The table reports the average AUC measured on the titration family along with their standard deviation (±). Ref. Branches indicate which combination of non-downsampled bootstrap support values (as estimated on 200 columns) were used to select the reference branches as branches having a support of 80 or more among the replicates of the considered methods. Bootstrap Combination indicates which bootstrap replicates (estimated on 25 columns) were combined (arithmetic mean) to estimate the bootstrap support values. The average AUC values indicate the discriminative capacity of the bootstrap combination between reference and non-reference branches as estimated on the original ML trees. Readouts in bold indicate bootstrap combinations featuring IMD. Number of Datasets indicates the number of discriminative datasets featuring both reference and non-reference branches that were used for the AUC analysis.

Quick links

Search