Table 3 Peptide sequence distributions. Peptides are divided into the two classes of binding orientation (I and II). The peptide proportion is reported in the second column. The third and fourth columns contain the number of binders and non-binders, respectively. The fifth column describes class I and class II SH3 domains, with the corresponding proportions of binders and non-binders listed in the last two columns, respectively. The latter information characterizes the domain-specific datasets used to train and test the corresponding domain-specific neural networks. The percentage of binders (3^rd and 6^th columns) highlights the critical unbalancing and attains acceptable levels only in the two class-specific datasets and in three class II domains in the domain-specific datasets.

From: A neural strategy for the inference of SH3 domain-peptide interaction specificity