Pripper: prediction of caspase cleavage sites from whole proteomes

BMC Bioinformatics

Table 2 Quality measures of trained classifiers and comparison with publicly available caspase cut site prediction models.

Classifier	ACC	PRC	FDR	SPC	MCC
SVM-6-4	87,4%	85,3%	14,7%	84,4%	74,8%
RF-12-12	85,7%	82,5%	17,5%	82,5%	71,7%
J48-4-2	81,6%	79,7%	20,3%	78,3%	63,3%
Vote	96,8%	100%	0,0%	100%	93,7%
PeptideCutter	50,8%	63,0%	37,0%	97,7%	4,6%
GraBCas	67,7%	67,6%	32,4%	67,5%	35,4%
CASVM P4-P1	62,3%	83,0%	17%	93,7%	31,6%
CASVM P4-P2'	72,7%	73,1%	26,9%	73,6%	45,4%
CASVMP14-P10'	83,1%	81,6%	18,4%	80,1%	66,2%

Peptide cutter, GraBCas and CASVM were all tested with the complete training data. The values obtained from SVM-6-4, J48-4-2 and RF-12-12 represent the values obtained from the test sequences of the training with the leave-one-out method. The default cut off value 1.2 was used in GraBCas. Classifiers RF-12-12 and SVM-6-4 were used in the Vote classifier. Abbreviations: ACC, accuracy; PRC, precision; FDR, false discovery rate; SPC, specificity; MCC, Matthews correlation coefficient.

ISSN: 1471-2105