Table 3.
Comparison of generalized classification across sources
|
Data source |
Classification method |
||
|---|---|---|---|
| SVM(RBF) | Ridge regr. | kNN | |
| AmiGO |
0.894 |
0.907 |
0.867 |
| BioCyc |
0.698 |
0.687 |
0.679 |
| CDD |
0.729 |
0.760 |
0.755 |
| GenNav |
0.940 |
0.935 |
0.878 |
| InterPro |
0.846 |
0.804 |
0.832 |
| Kegg |
0.733 |
0.778 |
0.779 |
| Kegg (pathways) |
0.740 |
0.739 |
0.717 |
| Pdb |
0.740 |
0.737 |
0.710 |
| TigrFam | 0.688 | 0.702 | 0.704 |
Results by source and method for predicting virulent and non-virulent bacterial proteins given AUC. The best performer, GenNav was run with a Gaussian kernel whose σ = 1.0 and regularization cost C = 1.0. For each method, the best performing classification approach is bolded.