Fig. 9.

Confusion matrix related to the predictions of nRC tool over the training dataset (6320 sequences). It has been obtained putting together the single confusion matrices produced at the end of each fold during the testing procedure. Classification of Intron gpI and Intron gpII classes gave the best result; classification of IRES, miRNA and HACA-box classes gave the worst results. In red we highlighted some situations, some of which discussed in Discussion section, that will need further investigation