Detecting Protein-Protein Interactions with a Novel Matrix-Based Protein Sequence Representation and Support Vector Machines

Table 3

Comparing the prediction performance by the proposed method and amino acid dipeptide composition method on the yeast dataset.

Methods	Kernel	Mean/std.	Testing
			ACC	SN	SP	PPV	NPV		MCC	AUC
The proposed method	Sigmoid	Mean	0.8734	0.8379	0.9092	0.9032	0.8474	0.8693	0.7784	0.9385
		Variance	0.0073	0.0093	0.0078	0.0087	0.0063	0.0088	0.0111	0.0071
	Gaussian	Mean	0.9006	0.8574	0.9437	0.9384	0.8689	0.8961	0.8203	0.9528
		Variance	0.0064	0.0094	0.0095	0.0098	0.0048	0.0076	0.0103	0.0064
	Polynomial	Mean	0.8963	0.8517	0.9408	0.9351	0.8639	0.8915	0.8134	0.9506
		Variance	0.0079	0.0072	0.0112	0.0118	0.0050	0.0085	0.0124	0.0061
	Linear	Mean	0.8642	0.8267	0.9016	0.8938	0.8389	0.8589	0.7646	0.9238
		Variance	0.0048	0.0098	0.0114	0.0103	0.0073	0.0052	0.0068	0.0038
AADC method	Sigmoid	Mean	0.6776	0.6726	0.6825	0.6792	0.6760	0.6758	0.5630	0.7343
		Variance	0.0088	0.0194	0.0098	0.0107	0.0136	0.0133	0.0062	0.0129
	Gaussian	Mean	0.8654	0.8349	0.8959	0.8892	0.8443	0.8612	0.7666	0.9292
		Variance	0.0065	0.0104	0.0047	0.0041	0.0119	0.0058	0.0095	0.0087
	Polynomial	Mean	0.8514	0.8196	0.8833	0.8754	0.8305	0.8465	0.7465	0.7540
		Variance	0.0063	0.0144	0.0078	0.0072	0.0110	0.0077	0.0090	0.3751
	Linear	Mean	0.8409	0.8150	0.8668	0.8597	0.8240	0.8367	0.7320	0.9021
		Variance	0.0060	0.0050	0.0146	0.0128	0.0070	0.0049	0.0080	0.0030