APIS: accurate prediction of hot spots in protein interfaces by combining protrusion index with solvent accessibility

Table 4 Evaluation of the hot spot prediction using different machine learning classifiers based on the RcsASA feature

Classifier	Dataset	Specificity	Recall	Precision	Accuracy	F1	TP	TN	FP	FN
SVM	Training set	0.79	0.74	0.71	0.77	0.72	46	73	19	16
	Test set	0.66	0.67	0.46	0.66	0.55	26	58	30	13
Bayes Net	Training set	0.79	0.56	0.65	0.70	0.60	35	73	19	27
	Test set	0.85	0.28	0.46	0.68	0.35	11	75	13	28
Naïve Bayes	Training set	0.75	0.81	0.68	0.77	0.74	50	69	23	12
	Test set	0.58	0.72	0.43	0.62	0.54	28	51	37	11
RBF Network	Training set	0.85	0.63	0.74	0.76	0.67	39	78	14	23
	Test set	0.76	0.62	0.53	0.72	0.57	24	67	21	15
Decision Tree (J48)	Training set	0.87	0.53	0.73	0.73	0.62	33	80	12	29
	Test set	0.84	0.28	0.44	0.67	0.34	11	74	14	28
Decision Table	Training set	0.79	0.56	0.65	0.70	0.60	35	73	19	27
	Test set	0.85	0.28	0.46	0.68	0.35	11	75	13	28

ISSN: 1471-2105