Predicting Protein Interactions Using a Deep Learning Method-Stacked Sparse Autoencoder Combined with a Probabilistic Classification Vector Machine

Table 4

The prediction performance comparison of PCVM with SVM.

Model	Testing set	Accu (%)	Sens (%)	Prec (%)	MCC (%)
PCVM	1	96.83	97.37	96.44	93.85
	2	96.33	97.33	95.22	92.93
	3	96.33	96.86	96.02	92.93
	4	96.60	96.85	96.33	93.44
	5	96.64	97.75	95.19	93.11
	Average	96.55 ± 0.2	97.23 ± 0.3	95.84 ± 0.5	93.25 ± 0.3
SVM	1	94.46	93.68	95.36	89.53
	2	93.70	90.32	96.46	88.13
	3	93.92	92.49	95.49	88.58
	4	92.76	91.99	93.33	86.56
	5	93.53	92.99	93.92	87.89
	Average	93.67 ± 0.6	92.29 ± 1.2	94.91 ± 1.2	88.13 ± 1.0