Predicting phosphorylation sites using machine learning by integrating the sequence, structure, and functional information of proteins

Table 5 Performance comparison with individual and combined feature encoding schemes for pY site prediction on the independent dataset

Attributes	Methods	Accuracy (%)	Sensitivity (%)	Specificity (%)	Precision (%)	F-measure (%)	MCC	AUC
Physiochemical property	RF	77.19	77.5	46	99.4	87.1	0.05	0.65
Physiochemical property	SVM	79.21	79.6	30.2	99.2	88.4	0.02	0.54
Structure	RF	99.3	100	79.4	99.3	99.7	0.43	0.95
Structure	SVM	96.08	96.4	63.5	99.7	98	0.27	0.79
Sequence	RF	69.74	70	41.3	99	82	0.02	0.59
Sequence	SVM	99	99.8	11.1	99.2	99.5	0.18	0.55
Functional features	RF	98.09	98.6	39.7	99.5	99.0	0.27	0.70
Functional features	SVM	97.73	98.2	39.7	99.5	98.9	0.24	0.69
Functional annotation	RF	95.51	96.1	31.7	99.4	97.7	0.12	0.68
Functional annotation	SVM	95.24	95.8	31.7	99.4	97.6	0.12	0.63
Combined	RF	99.42	100	63.5	99.5	99.7	0.57	0.99
Combined	SVM	99.46	99.7	23.8	99.7	99.7	0.71	0.87

ISSN: 1479-5876