Computational screening of structurally related impurities for predicted DPP-4 binding affinity

Process-related impurities and degradation products of dipeptidyl peptidase-4 (DPP-4) inhibitors may retain predicted DPP-4 binding affinity despite being present at trace levels, representing a potential challenge for pharmaceutical quality and safety assessment. The aim of this study was to computationally evaluate DPP-4 docking profiles of structurally related impurities of clinically used gliptins and identify molecular features associated with docking-score variation. A comprehensive dataset comprising 79 process-related impurities and degradation products together with 14 parent DPP-4 inhibitors (93 compounds in total) was investigated using molecular docking with AutoDock Vina. Molecular descriptors calculated using the OCHEM platform were subsequently analyzed by genetic algorithm-multiple linear regression (GA-MLR) to identify molecular features associated with docking-score variation. Docking scores ranged from -12.0 to -4.2 kcal mol-1, with several linagliptinand alogliptin-related impurities exhibiting docking scores comparable to or more favorable than those of their corresponding parent drugs. The developed GA-MLR model demonstrated satisfactory goodness-of-fit and internal validation performance (R² = 0.936, Q² LOO = 0.925, CCC = 0.967) and identified molecular topology, electronic organization, steric complementarity, and charge distribution as molecular features associated with DPP-4 docking behaviour. These findings suggest that preservation of key pharmacophore features, rather than structural similarity alone, contributes to favorable predicted DPP-4 recognition by pharmaceutical impurities. Several structurally related impurities, particularly those associated with linagliptin and alogliptin, retained favorable DPP-4 docking profiles comparable to those of the corresponding parent compounds, supporting their prioritization for subsequent experimental evaluation.