ROC curves for the different models. ROC curves obtained upon classification of the test human and murine databases, using the human learning database to learn the MG distribution, and to calculate the average distance for the “Tall” and “Tn” methods. Panel (a): ROC curves obtained using the full VHVL chain for classification, with or without CDR regions. Panel (b): ROC curves obtained using the VH or VL chain separately for classification. “Corr. off” indicates that correlations between residues have been removed in the MG model, “FR” refers to the curve obtained removing the CDRs and keeping just the framework regions, “h-m” indicates that classification is performed using both human and murine learning datasets as reference (see Methods).