Comparison of the proposed BEST method with existing B-cell epitope predictors on the SEQ149 dataset.

The methods are sorted alphabetically within each category. We evaluate significance of differences between BEST16 (BEST10) and the other methods. We compare the corresponding AUC values in 10 paired results based on 100 random selected chains from the SEQ194 dataset using paired t-test; +/– mean that BEST16 (BEST10) are significantly better/worse that another method at p-value aresults from [34].bresults from the Epitopia web server at http://epitopia.tau.ac.il/.cresults from the ABCPred web server http://www.imtech.res.in/raghava/abcpred/.dresults from the BayesB web server at http://www.immunopred.org/bayesb/index.html.eresults from the CBTOPE web server at http://www.imtech.res.in/raghava/cbtope/.fresults from the COBEpro web server at http://scratch.proteomics.ics.uci.edu/.gresults generated using BEST method, which is based on the SVM model (C = 1.0 and gamma = 0.001953) with 84 features generated with the Filtered40_BCPREDFrag dataset and the distance scheme with k = 16 (BEST16) and with k = 10 (BEST10).