Figure 3. Histograms showing type distributions of the optimal features selected from the 4 datasets, respectively.

The optimal features were selected when the MCC value reached its maximum trained on 17 length peptides in the training set of the 4 datasets, respectively.