Prediction of B-cell Linear Epitopes with a Combination of Support Vector Machine Classification and Amino Acid Propensity Identification
Figure 1
The design of LEPS. (a) Step 1(a): primitive epitope candidates with globally and locally high antigenicity were extracted by calculating weighting coefficients for various physicochemical propensities of each amino acid. After the filtering process with the SVM classifier (step 2(a)), predicted epitopes were highlighted (step 3(a)) in the query sequence and the simulated structure. (b) Step 1(b): 1230 experimentally verified epitopes and 872 non-epitopes were analyzed to determine the statistical characteristics of AASs. Step 2(b): subsequently, epitope indexes of 872 epitopes and 872 non-epitopes were used to train the SVM model to predict candidate epitopes based on the statistical characteristics defined in step 1(b).
