Table 3.

Average of sensitivity, specificity, precision and standard deviations for OTU-level identification over 500 replications

Method	Graph type	Sensitivity (SD)	Specificity (SD)	Precision (SD)
STB	Random	0.2815 (0.0609)	0.9979 (0.0015)	0.9073 (0.0438)
STC		0.1377 (0.0724)	0.9993 (0.0011)	0.9528 (0.0637)
ZIBR		0.1463 (0.0486)	1 (0)	0.5083 (0.2771)
ANCOM-BC		0.7225 (0.0152)	0.6084 (0.0078)	0.1702 (0.0048)
Hu et al. [14]		0.0047 (0.0005)	0.9993 (0.0006)	0.4520 (0.4915)
STB	Hub	0.3067 (0.0425)	0.9986 (0.0009)	0.9218 (0.0557)
STC		0.1183 (0.0310)	0.9991 (0.0017)	0.9198 (0.1473)
ZIBR		0.1439 (0.0910)	0.9999 (0.0012)	0.4827 (0.2760)
ANCOM-BC		0.7927 (0.0154)	0.5204 (0.0168)	0.1465 (0.0032)
Hu et al. [14]		0.0025 (0.0046)	0.9991 (0.0005)	0.2194 (0.4068)
STB	Cluster	0.3255 (0.0513)	0.9974 (0.0018)	0.9065 (0.0533)
STC		0.1339 (0.0425)	0.9998 (0.0004)	0.9851 (0.0354)
ZIBR		0.1160 (0.0811)	0.9993 (0.0033)	0.3760 (0.2261)
ANCOM-BC		0.8107 (0.0155)	0.5789 (0.0228)	0.1749 (0.0073)
Hu et al. [14]		0.0034 (0.0048)	0.9992 (0.0005)	0.3280 (0.4667)

Different graph types are set to reflect various correlation structures among OTUs. Random graph structure indicates that OTUs are correlated with each other randomly. The hub and cluster graphs capture some aspects of biological networks, such as highly connected nodes and community structure. More details can be found in Osborne et al. [20]