Prediction of activity and selectivity profiles of human Carbonic Anhydrase inhibitors using machine learning classification models

Table 5 Accuracy, MCC, precision and recall values for best models in the external validation phase

Probability score	Overall	0.6	0.7	0.8	0.9	1
hCA II, N = 350
TP	782	151	223	131	150	127
FN	741	279	228	167	48	18
FP	204	70	103	17	5	9
TN	683	185	192	110	125	70
Accuracy	0.61	0.49	0.56	0.57	0.84	0.88
MCC	0.28	0.08	0.14	0.29	0.70	0.75
Precision	0.79	0.68	0.68	0.89	0.97	0.93
Recall	0.51	0.35	0.49	0.44	0.76	0.88
hCA IX, N = 450
TP	326	46	103	75	59	43
FN	110	46	29	15	13	6
FP	61	15	20	11	8	7
TN	212	38	58	32	46	37
Accuracy	0.76	0.58	0.77	0.80	0.83	0.86
MCC	0.51	0.21	0.51	0.57	0.67	0.72
Precision	0.84	0.75	0.84	0.87	0.88	0.86
Recall	0.75	0.50	0.78	0.83	0.82	0.88
hCA XII, N = 150
TP	524	88	183	134	55	64
FN	271	94	116	35	20	6
FP	32	7	13	7	4	1
TN	63	20	12	4	19	8
Accuracy	0.66	0.52	0.60	0.77	0.76	0.91
MCC	0.21	0.15	0.05	0.09	0.48	0.67
Precision	0.94	0.93	0.93	0.95	0.93	0.98
Recall	0.66	0.48	0.61	0.79	0.73	0.91

Results are averaged on 10 independent experiments and then rounded to the nearest integer value. Performance metrics are reported in italics
TP true positives, FN false negatives, FP false positives and TN true negatives

ISSN: 1758-2946