Table 3.

Mean performance characteristics of the taxonomy- and phenotype-based CD classifiers over 10 classification iterations.

Strategy/dataset	Taxonomy-based classifier			Phenotypes-based classifier
	Mean sensitivity	Mean specificity	Mean AUC	Mean sensitivity	Mean specificity	Mean AUC
Single:CHN	0.7867	0.7857	0.8463	0.7333	0.7571	0.8463
Single:ESP	0.7900	0.9179	0.9511	0.7700	0.9036	0.915
Single:NLD	0.7755	0.9161	0.9344	0.6918	0.9020	0.8920
L1O:CHN	0.9440	0.4358	0.7985	0.850	0.6164	0.7725
L1O:ESP	0.9353	0.8165	0.9484	0.8176	0.8516	0.8895
L1O:NLD	0.4252	0.9871	0.8567	0.3534	0.9899	0.7632
Mixed: all datasets	0.7548	0.8083	0.8736	0.7065	0.8350	0.8339
Mean across all variants	0.7731	0.8096	0.8870	0.7032	0.8365	0.8446

In the L1O strategy description, the name of the dataset corresponds to the testing dataset (for example, the “L1O:CHN” description means that the classifier was trained on the ESP and NLD datasets and tested on the CHN dataset). Cell color reflects the characteristics' values (greater values correspond to darker colors).