MOTIVATION: Many secretory peptides are synthesized as inactive precursors that must undergo post-translational processing to become biologically active peptides. Attempts to predict natural peptides are limited by the low performance of proteolytic site predictors and by the high combinatorial complexity of pairing such sites. To overcome these limitations, we analyzed the site-wise evolutionary mutation rates of peptide hormone precursors, calculated using the Rate4Site algorithm. RESULTS: Our analysis revealed that within their precursors, peptide residues are significantly more conserved than the pro-peptide residues. This disparity enables the prediction of peptides with a precision of ∼60% at a recall of 40% [receiver-operating characteristic curve (ROC) AUC 0.79]. Subsequently, combining the Rate4Site score with additional features and training a Random Forest classifier enable the prediction of natural peptides hidden within secreted human proteins at a precision of ∼90% at a recall of 50% (ROC AUC 0.96). The high performance of our method allows it to be applied to full secretomes and to predict naturally occurring active peptides. Our prediction on Homo sapiens revealed several putative peptides in the human secretome that are currently unannotated. Furthermore, the unique expression of some of these peptides implies a potential hormone function, including peptides that are highly expressed in endocrine glands. AVAILABILITY AND IMPLEMENTATION: A pseudocode is available in the SUPPLEMENTARY INFORMATION. CONTACT: [email protected] or [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Many secretory proteins are synthesized as inactive precursors that must undergo post-translational proteolysis in order to mature and become active. In the current study, we address the challenge of sequence-based discovery of proteolytic sites in secreted proteins using machine learning. RESULTS: The results revealed that only half of the extracellular proteolytic sites are currently annotated, leaving over 3600 unannotated ones. Furthermore, we have found that only 6% of the unannotated sites are similar to known proteolytic sites, whereas the remaining 94% do not share significant similarity with any annotated proteolytic site. The computational challenges in these two cases are very different. While the precision in detecting the former group is close to perfect, only a mere 22% of the latter group were detected with a precision of 80%. The applicability of the classifier is demonstrated through members of the FGF family, in which we verified the conservation of physiologically-relevant proteolytic sites in homologous proteins.