Skip to main content

Table 7 Performance of 3-way classification using SVM classifier (feature length 20)

From: Plus ça change – evolutionary sequence divergence predicts protein subcellular localization signals

 

Divergence

Classical features

Combination

 

AUC

MCC

AUC

MCC

AUC

MCC

MTS

0.67±0.03

0.36±0.06

0 . 8 9±0.02

0.80±0.02

0 . 8 9±0.01

0 . 8 1±0.02

SP

0.50±0.00

0.00±0.00

0.97±0.03

0.92±0.07

0 . 9 8±0.03

0 . 9 7±0.04

N-signal-free

0.66±0.02

0.36±0.03

0 . 9 0±0.01

0.81±0.02

0 . 9 0±0.01

0 . 8 3±0.02

% accuracy

70.82±1.61

91.49±1.26

9 2 . 2 3±1.25

  1. The 5-fold cross-validation performance of an SVM classifier using: divergence features only, classical features only, and the two combined; is shown for three-way classification on our entire yeast curated ortholog dataset. Classical features are calculated from N-terminal 20 amino acids.