Subcellular location prediction of proteins using support vector machines with alignment of block sequences utilizing amino acid composition

Table 6 Comparison of predictive accuracy of our methods for plant proteins in the TargetP data set. "cTP", "mTP", "SP", and "other" indicate proteins destined for chloroplast, mitochondria, secretory pathway, and other locations (nucleus and cytosol), respectively.

Predictor	Location	Sensitivity	Specificity	MCC	Average MCC	Overall Accuracy
Our method (SVM with specifying posconstraint and negconstraint)	cTP	0.8440	0.9015	0.8507	0.8655	0.9096
	mTP	0.9348	0.9125	0.8735
	SP	0.9665	0.9319	0.9282
	other	0.8148	0.8684	0.8095
Our method (SVM without specifying posconstraint and negconstraint)	cTP	0.7518	0.9550	0.8249	0.8525	0.8989
	mTP	0.9592	0.8506	0.8363
	SP	0.9554	0.9554	0.9375
	other	0.7963	0.8897	0.8111
Nearest Neighbor Classifier (with our similarity measure)	cTP	0.7447	0.7664	0.7131	0.7255	0.8128
	mTP	0.8750	0.7854	0.7098
	SP	0.8848	0.9015	0.8508
	other	0.6111	0.7674	0.6284

Note: In the first method, all parameters of Table 2 are given. In the second and third methods, gap penalty = 0.4, γ = 2.5, c = 10, w = 20 are given although posconstraint and negconstraint are not specified.

ISSN: 1471-2105