Peptide-protein interactions play a key role for many cellular and metabolic processes involved in the onset of largely spread diseases such as cancer and neurodegenerative pathologies. Despite the progress in the structural characterization of peptide-protein interfaces, the in-depth knowledge of the molecular details behind their interactions is still a daunting task. Here, we present the first comprehensive in silico morphological and energetic study of peptide binding sites by focusing on both peptide and protein standpoints. Starting from the PixelDB database, a nonredundant benchmark collection of high-quality 3D crystallographic structures of peptide-protein complexes, a classification analysis of the most representative categories based on the nature of each cocrystallized peptide has been carried out. Several interpretable geometrical and energetic descriptors have been computed both from peptide and target protein sides in the attempt to unveil physicochemical and structural causative correlations. Finally, we investigated the most frequent peptide-protein residue pairs at the binding interface and made extensive energetic analyses, based on GRID MIFs, with the aim to study the peptide affinity-enhancing interactions to be further exploited in rational drug design strategies.