A comprehensive statistical analysis of the geometric parameters of peptide chains in a reduced dataset of protein-peptide complexes in PDB is presented. The angular variables describing the backbone conformations of amino acid residues in peptide chains shed insights into the conformational preferences of peptide residues interacting with protein partners. Non-parametric statistical approaches are employed to evaluate the interrelationships and associations in structural variables. Grouping of residues based on their structure into chemical classes reveals characteristic trends in parameter relationships. A comparison of canonical amino acid residues in free peptide structures in Cambridge Structural Database (CSD) with identical residues in PDB complexes, suggests that the information can be integrated from both the structural repositories enabling efficient and accurate modelling of biologically active peptides. This article is protected by copyright. All rights reserved.