Table 7.

Performance comparison of the general purpose and ligand-specific models of the TargetSeq method on the dataset of the 9 ligands by five-fold cross-validation

Ligand Type	Model Type	Accuracy (%)	Sensitivity (%)	Specificity (%)	MCC
CU	General	86.98	79.62	87.09	0.22
	Specific	99.02	51.40	99.69	0.59
FE	General	90.14	85.02	90.22	0.29
	Specific	98.83	53.54	99.52	0.57
FE2	General	90.67	90.89	90.67	0.30
	Specific	99.20	51.36	99.81	0.63
ZN	General	88.50	74.29	88.66	0.27
	Specific	99.01	41.78	99.68	0.50
SO4	General	85.85	55.29	86.50	0.17
	Specific	97.79	10.07	99.66	0.19
PO4	General	86.38	71.73	86.66	0.23
	Specific	97.29	27.86	98.63	0.27
ATP	General	87.46	71.88	88.02	0.32
	Specific	96.73	48.09	98.48	0.49
FMN	General	88.24	76.68	88.83	0.40
	Specific	96.39	66.56	97.92	0.62
HEME	General	86.18	73.85	87.21	0.43
	Specific	93.63	58.24	96.61	0.55