Table 2

The 10-fold cross-validation results by SVM classifiers based on different features and those by ensemble classifiers for 10 locations with respect to AUC scores, where the ensemble classifiers are the optimal combinations of different SVM classifiers trained on features without strikethrough. The numbers with strikethrough indicate the corresponding classifier was not used in the ensemble classifier. The numbers within the brackets denote the corresponding Gap.

Localization

threAA

N-term

C-term

Gap1

Gap2

Ensemble


Extracellular

0.909

0.891

0.812

0.943(13)

0.943(7)

0.950

Cytoplasm

0.652

0.653

0.643

0.637(15)

0.630(5)

0.738

ER

0.789

0.681

0.703

0.811(11)

0.806(13)

0.827

Golgi apparatus

0.748

0.791

0.764

0.732(7)

0.729(5)

0.848

Nucleus

0.688

0.630

0.615

0.709(13)

0.707(15)

0.721

Mitochondrion

0.722

0.781

0.665

0.805(15)

0.802(13)

0.833

Peroxisome

0.808

0.797

0.777

0.816(15)

0.815(13)

0.882

Endosome

0.705

0.787

0.795

0.860(13)

0.848(11)

0.895

Vacuole

0.730

0.749

0.758

0.748(15)

0.746(9)

0.820

Cell membrane

0.798

0.792

0.786

0.801(11)

0.800(9)

0.837


Sun et al. BMC Systems Biology 2010 4(Suppl 2):S12   doi:10.1186/1752-0509-4-S2-S12

Open Data