Predicting Interactions between Virus and Host Proteins Using Repeat Patterns and Composition of Amino Acids

Table 4

Results of testing our SVM model with different partitions of a protein sequence on three datasets.

Our dataset
F3	Sn (%)	Sp (%)	Acc (%)	PPV (%)	NPV (%)	MCC	AUC
5 partitions	88.24	97.34	92.79	97.07	89.22	0.859	0.963
7 partitions	88.24	97.96	93.10	97.74	89.29	0.866	0.965
9 partitions	89.19	96.08	92.63	95.79	89.88	0.855	0.962
DeNovo dataset
F3	Sn (%)	Sp (%)	Acc (%)	PPV (%)	NPV (%)	MCC	AUC
5 partitions	86.35	86.59	86.47	86.56	86.39	0.729	0.926
7 partitions	83.60	81.18	82.41	82.30	82.54	0.648	0.907
9 partitions	84.27	79.53	81.95	81.17	82.84	0.639	0.902
Barman dataset
F3	Sn (%)	Sp (%)	Acc (%)	PPV (%)	NPV (%)	MCC	AUC
5 partitions	73.72	83.48	78.60	81.69	76.06	0.575	0.847
7 partitions	78.55	78.55	78.55	78.55	78.55	0.571	0.858
9 partitions	78.16	79.81	78.99	79.47	78.52	0.580	0.860
Average of the above three results
F3	Sn (%)	Sp (%)	Acc (%)	PPV (%)	NPV (%)	MCC	AUC
5 partitions	82.77	89.14	85.95	88.44	83.89	0.721	0.912
7 partitions	83.46	85.90	84.69	86.20	83.46	0.695	0.910
9 partitions	83.87	85.14	84.52	85.48	83.75	0.691	0.908

All the results were obtained by commonly using SAR for features F1 and F2.