Figure 4

Performance of different variations of our method during the design of EASE-AA. The unseen-protein cross-validation performance [Matthews correlation coefficient (MCC) and the area under the ROC curve (AUC)] is shown for four different variations of our method. The difference between AAP+D+ASA+EASE and EASE-AA is the removal of the 140 input attributes defining the identities of the deleted, introduced, and neighbouring amino acids.