Prediction of cis/trans isomerization in proteins using PSI-BLAST profiles and secondary structure information

Table 6 Comparison of predictive performance with other methods.

Methods	Prediction accuracy (%)		Dataset used	Prediction performance evaluation method
	Q ₂	MCC
Statistical pattern^a	72.7	-	242 Xaa-Pro bonds	self-consistency
COPS^b	63.6	-	8584 proteins	10-fold cross-validation
SVM single sequence^c	69.8	-	2193 proteins	independence test
SVM single sequence^d	76.6	0.53	2193 proteins	Jack-knife
SVM single sequence^e	62.8	0.26	2424 proteins	5-fold cross-validation
SVM PSI-BLAST^f	69.8	0.40	2424 proteins	5-fold cross-validation
SVM PSIPRED^g	63.6	0.27	2424 proteins	5-fold cross-validation
SVM PSI-BLAST and PSIPRED^h	71.5	0.43	2424 proteins	5-fold cross-validation

^aPrediction accuracy reported by Frömmel and Preissner [16]. The result cannot be determined from the paper.
^bPrediction accuracy estimated based on the average statistical results of COPS [18].
^cPrediction accuracy using independence test reported by Wang et al [17].
^dPrediction accuracy using jack-knife test reported by Wang et al [17].
^ePrediction accuracy of SVM based on single sequence encoding scheme using our dataset.
^fPrediction accuracy of SVM based on PSI-BLAST encoding scheme using our dataset.
^gPrediction accuracy of SVM based on PSIPRED encoding scheme using our dataset.
^hPrediction accuracy of SVM based on PSI-BLAST and PSIPRED encoding scheme using our dataset.

ISSN: 1471-2105