Skip to main content

Table 3 Performance1 of the best predictors for the three different prediction schemes

From: 'Unite and conquer': enhanced prediction of protein subcellular localization by integrating multiple specialized tools

Classes2

Individual tool (PASUB)

Combination of tools by voting3

Decision tree classifier (STACK-mem-DT)

  

TPR

FPR

ACC

TPR

FPR

ACC

TPR

FPR

ACC

Yeast

Mit

0.74

0.05

0.69

0.75

0.02

0.84

0.92

0.05

0.95

 

Non

0.65

0.06

 

0.99

0.20

 

0.97

0.05

 

Arabidopsis

Mit

0.75

0.09

0.81

0.67

0.07

0.88

0.87

0.12

0.94

 

Non

0.83

0.05

 

0.95

0.09

 

0.96

0.04

 

Human

Mit

0.87

0.09

0.68

0.88

0.01

0.97

0.90

0.02

0.99

 

Non

0.65

0.02

 

0.98

0.02

 

0.99

0.01

 
  1. 1 TPR: true positive rate; FPR: false positive rate; ACC: accuracy (all correctly predicted instances/all instances)
  2. 2 Mit: mitochondrial proteins; Non: proteins of other subcellular locations
  3. 3 The best combination of tools is pTARGET+PASUB+CELLO for yeast data, PASUB+MitoPort+CELLO for Arabidopsis data, and pTARGET +SherLoc+ PASUB for human data