Table 1.
Synoptic summary of the main characteristics of the six programs tested.
| Tool | Category | PVC selection criterion | Advantage | Disadvantage | Number of studies that used the tool |
|---|---|---|---|---|---|
| NERVE | Decision-tree | No cytoplasmatic protein < 2 transmembrane helices High adhesin probability No homology with human proteins | Input and output data are automatically structured in a database | Not updated | 4 |
| VaxiJen | Machine-learning | Output probability greater than a cut-off (0.5) | Very fast Graphical interface | Fixed training datasets (100 known bacterial antigens, 100 putative non-antigens) | 20 |
| Vaxign | Decision-tree | No cytoplasmatic protein < 2 transmembrane helices High adhesin probability No homology with human and mouse proteins | Regularly maintained Easy to use and intuitive | Download of the results is limited to 500 proteins | 18 |
| Jenner-predict | Decision-tree | No cytoplasmatic protein < 2 transmembrane helices Presence of Pfam domains involved in host-pathogen interaction and pathogenesis | Upload and download of large datasets | Temporarily unavailable | 1 |
| Bowman-Heinson | Machine-learning | Output probability greater than a cut-off (0.5) | Larger training set (200 known bacterial antigens, 200 putative non-antigens) | Annotation tools for eukaryotes used for bacterial proteins Pipeline not delivered | 0 |
| VacSol | Decision-tree | No cytoplasmatic protein < 2 transmembrane helices No homology with human proteins Essential gene Virulence factor | User-friendly interface | Too restrictive | 0 |