Table 3.
Comparison of available web-based tools for predicting linear B cell epitopes.
| Tool | Algorithm | Training dataset | Validation | URL | Reference | |
|---|---|---|---|---|---|---|
| B cell epitopes | Non-B cell epitopes | |||||
| BepiBlast | BLAST | 62,730 | – | X, I | http://imath.med.ucm.es/bepiblast/ | – |
| Bceps | Support vector machine | 555 | 555 (a) | X, I, E | http://imath.med.ucm.es/bceps/ | 18 |
| BepiPred 2.0a | Random forest | 3542 | 36,785 | X, I, E | https://services.healthtech.dtu.dk/service.php?BepiPred-2.0 | 20 |
| LBtopeb | Support vector machine | 14,876 | 23,321 (b) | X, I | https://webs.iiitd.edu.in/raghava/lbtope/ | 17 |
| IBCE-EL | Random tree with boosting | 4440 | 5485 (b) | X, I | http://www.thegleelab.org/iBCE-EL/ | 28 |
| DLBEpitope | Deep neural network | 22,012 | 201,563 (b) | X, I | http://ccb1.bmi.ac.cn:81/dlbepitope/index.php? | 15 |
| ILBE | Random Forest | 4440 | 5485 (b) | X, I | http://kurata14.bio.kyutech.ac.jp/iLBE/ | 41 |
| ABCPred | Neural network | 700 | 700 (a) | X, I | https://webs.iiitd.edu.in/raghava/abcpred/ | 14 |
| BCPREDS | Support vector machine | 701 | 701 (a) | X, I, E | http://ailab.ist.psu.edu/bcpred/ | 32 |
| SVMtrip | Support vector machine | 4925 | 4925 (b) | X | http://sysbio.unl.edu/SVMTriP/prediction.php | 16 |
For each tool, table reports the underlying algorithm; the number of B and non-B cell epitopes for model building; the method used for validation (X: cross-validation; I: independent dataset; E: case example); the URL of the tool and the reference. The letter between parenthesis indicates the type of non-B cell epitopes in the training dataset: a, random peptide sequences; b, peptide sequences with reported negative B cell epitope assays. aFor BepiPred, B and non-B cell epitope figures correspond to antigen residues that in the tertiary structure of antibody-antigen complexes contact the antibody or not, respectively. bData for default model in LBtope.