Fig. 1
From: Immunoinformatics and epitope prediction in the age of genomic medicine
Generating predictions from data. a Evaluation of the predictor using cross-validation: first the data-set is split into k-folds (k = 5). Next, five predictors are trained on four folds and validated on the one left out. Evaluation can be, for example, a receiver operating characteristic (ROC) curve analysis. Finally, an average ROC curve is generated. b Training of the final predictor: after evaluation, the final predictor is trained on the complete data-set