HLAB27Pred is a server designed for the prediction of HLA-B*2705 (MHC class I allele) based nanomer epitopic binding peptides. This server implements two (2) techniques for the purpose of prediction, viz. SVM and PSSM.
SVM (Support Vector Machine) based prediction are deployed by training a set of experimentally validataed nanomeric binding and non-binding peptides. The average accuracy of this tool is 85.05% with an average Kappa value of 0.7. The performance of the SVM predictions has been tested through 5 cross-validation. The specificity and sensitivity obtained during the development of this server is 84.54% and 85.57% respectively. Whereas average precision and average recall values were observed to be 84.69% and 85.57% respectively.
PSSM (Position Specific Scoring Matrix) method has also been implemented for the prediction of binding peptides. Scoring matrix for this purpose has been developed based on the validated nanomeric binding pepdites used for the training of SVM.
The binding data for training of SVM model and generating PSSM has been extracted from IEDB database. Where as non-binding epitopic data was obtained from swissprot proteins. This tool will be useful in Cellular immunology, Vaccine design, Immunodiagnostics, Immunotherapeutics and molecular understanding of autoimmune susceptibility.