A generalized approach to predicting protein-protein interactions between virus and host

BMC Genomics

Table 3 Results of 10–fold cross validation of SVM model on 12,157 PPIs between any host-virus PPIs with different ratios of positive to negative instances

P:N	Dataset	SN(%)	SP(%)	ACC(%)	PPV(%)	NPV(%)	MCC	AUC
	1	84.93	86.03	85.48	85.87	85.09	0.709	0.926
1:1	2	84.92	86.06	85.49	85.89	85.09	0.701	0.926
	3	85.36	85.92	85.64	85.84	85.44	0.712	0.925
	mean ± SD	85.07 ± 0.3	86.00 ± 0.1	85.54 ± 0.1	85.87 ± 0.0	85.21 ± 0.2	0.71 ± 0.0	0.93 ± 0.0
	1	78.91	91.17	87.08	81.72	89.64	0.707	0.923
1:2	2	78.29	91.03	86.78	81.36	89.34	0.700	0.921
	3	78.22	91.18	86.86	81.59	89.33	0.701	0.920
	mean ± SD	78.47 ± 0.4	91.13 ± 0.1	86.91 ± 0.2	81.56 ± 0.2	89.44 ± 0.2	0.70 ± 0.0	0.92 ± 0.0
	1	74.55	93.32	88.63	78.82	91.66	0.691	0.920
1:3	2	74.61	93.56	88.82	79.43	91.70	0.696	0.919
	3	74.62	93.41	88.72	79.07	91.69	0.693	0.920
	mean ± SD	74.59 ± 0.0	93.43 ± 0.1	88.72 ± 0.1	79.11 ± 0.3	91.68 ± 0.0	0.69 ± 0.0	0.92 ± 0.0

SN: sensitivity, SP: specificity, ACC: accuracy, PPV: positive predictive value, NPV: negative predictive value, MCC: Matthews correlation coefficient, AUC: the area under the ROC

ISSN: 1471-2164