SEPPA (Spatial Epitope Prediction of Protein Antigens) server is presented here as a tool for conformational B-cell epitope prediction. With 3D protein structure as input, each residue in the query protein will be given a score according to its neighborhood residues' information. Higher score corresponds to higher probability the residue to be involved in an epitope.

Citation:

Jing Sun, Di Wu, Tianlei Xu, Xiaojing Wang, Xiaolian Xu, Lin Tao, Y. X. Li, and Z. W. Cao
SEPPA: a computational server for spatial epitope prediction of protein antigens
Nucleic Acids Res., July 1, 2009; 37(suppl_2): W612 - W616.
[Abstract] [Full Text] [PDF]

Submission methods:

1. Submit the entry of an existing PDB ID.
2. Submit a protein structure file in your own disk. PDB format is required.
   
   

Threshold:

The default value of THRESHOLD is set at 1.80 to help to specify the epitope residues. Under the default threshold, a sensitivity of 0.568 was received while the specificity was 0.740 on SEPPA training dataset. As a reference, other thresholds with corresponding sensitivity/specificity/accuracy values are listed in Table 1.

Users can set different thresholds. Under a lower threshold, more residues will be included as predicted epitope residues. That always results in the increasement of the true positive rate and the false positive rate at the same time.

Figure 1. ROC curve of SEPPA. The ROC curve for training dataset.

Table 1. Performance of SEPPA. The sensitivity/specificity/accuracy with different thresholds. Threshold Sensitivity Specificity Accuracy 1.55 0.959 0.259 0.377 1.60 0.927 0.345 0.448 1.65 0.859 0.452 0.531 1.70 0.778 0.558 0.612 1.75 0.672 0.658 0.684 1.80 0.568 0.740 0.739 1.85 0.459 0.810 0.782 1.90 0.363 0.860 0.809 1.95 0.278 0.900 0.829 2.00 0.217 0.927 0.844

Output description:

Example: PDB ID: 1A14  chain ID: N

1. Part I: summary of prediction result
   ***********************************************************************************
   Chain: N
   Threshold: 1.80
   Number of total residues: 388
   Number of predicted epitope residues: 48
   ***********************************************************************************
   This part is a summary of the prediction result.
   
   
   
2. Part II: visualization of the prediction result
   ***********************************************************************************
   View 3D structure in Jmol
   Figure 2. Visualization of SEPPA prediction result.
   ***********************************************************************************
   The predicted epitope could be visualized with Jmol in different renderings. In the visualized structures, a RWB gradient is used for result display: tints from blue to red represent a rising antigenicity. Structure was initialized in a wireframe model (Figure 2. left). By selecting the "Highlighted epitope residues predicted" checkbox, Jmol changes the rendering so that the predicted epitope residues are solid spheres. Access the Jmol menu by clicking the word "Jmol" in the lower right corner of the structure display area.
   
   
   
3. Part III: a glance of the prediction result
   ***********************************************************************************
   
   ***********************************************************************************
   The prediction result is displayed in a table as above. Residues are listed sequentially. The predicted epitope residues are highlighted in yellow. The core residues are shown in lowercase.
   
   
   
4. Part IV: score file of the prediction result
   ***********************************************************************************
   

   N	82	ARG	1.67
   N	83	ASP	1.66
   N	84	PHE	1.64
   N	85	ASN	1.51
   N	86	ASN	1.54
   	....

   ***********************************************************************************
   The complete score file contains four columns, which are chain identifier, residue sequence number, residue name and SEPPA score.
   * The label "core" is used in the score column to represent a residue in the protein core.
   

Training dataset:

Non-redundant 84 antigens (from 82 immunological complexes) were collected from PDB as SEPPA training dataset.

Testing datasets:

There are 75, 140 and 72 antigens in DiscoTope training dataset, Epitome database and IEDB database. From these datasets, 119 antigens were retained as testing dataset, which are new to SEPPA training dataset.

ANNOUNCEMENT:

The residue solvent accessible areas are calculated with Naccess V2.1.1.
Hubbard SJ, Thornton JM. 'NACCESS', Computer Program. Department of Biochemistry and Molecular Biology, University College London. 1993.

Jmol has been implemented into SEPPA server.
Jmol: an open-source Java viewer for chemical structures in 3D.