Table 3.
Accuracy of prediction methods on sequence-unique data set
| Extracellular | Cytoplasmic | Nuclear | Mitochondrial | |||||
| Methoda | oLb | pLb | oL | pL | oL | pL | oL | pL |
| NNPSL | 62.6 | 30.3 | 43.6 | 50.2 | 65.1 | 72.2 | 67.3 | 35.5 |
| SubLoc | 72.9 | 52.7 | 66.9 | 59.7 | 82 | 76.1 | 65.0 | 50.2 |
| TargetP | 98.8 | 50.9 | – | – | – | – | 92.7 | 45.9 |
a Method used to predict localization. NNPSL: neural network-based tool for predicting subcellular localization based on amino acid composition (Reinhardt and Hubbard 1998); SubLoc: a support vector machine-based tool for predicting subcellular localization based on amino acid composition (Hua and Sun 2001); TargetP: neural network-based tool for large-scale subcellular localization prediction based on amino-terminal sequence information (Emanuelsson et al. 2000).
b Two-state accuracy (Eq. 5): oL = correctly predicted in localization L of all observed in L, pL = correctly predicted in L of all predicted in L. For instance, TargetP strongly overpredicts extracellular and mitochondrial proteins, therefore yielding high values for oL and low values for pL.