This study aims to identify and characterize, through in silico approaches, potential inhibitors of serine- and metallo-β-lactamases, enzymes directly associated with antimicrobial resistance (AMR). The ChEMBL database was screened to select molecules with reported bioactivity against TEM-1 (PDB ID: 1ERM) and KPC-2 (PDB ID: 6D15), both class A enzymes, and BcII (PDB ID: 2M5D), a class B enzyme. Based on IC₅₀ values, 200 molecules were selected for TEM-1, 89 for KPC-2, and 100 for BcII. The protonation states of the selected compounds at pH 7.4 were determined using MarvinSketch. Three-dimensional structures were built in Discovery Studio Visualizer and optimized using the semiempirical PM7 method implemented in MOPAC. Protocol validation was carried out by redocking the co-crystallized ligands, followed by molecular docking simulations of the selected compounds. Docking results were analyzed using dbCICA, which correlates binding affinity with interaction patterns in the enzyme active site. Post-processing of docking data enabled the identification of intermolecular interaction patterns that may assist the rational design of novel inhibitors. The most active ligands established multiple hydrogen bonds with conserved catalytic residues, including SER70, SER130, GLU166, ASN170, and LYS73. In contrast, less active compounds exhibited fewer interactions with these key residues. The dbCICA models showed limited predictive performance, with low coefficients of determination (r² = 0.24 for TEM-1, 0.37 for KPC-2, and 0.29 for BcII). Among the evaluated residues, GLU166 was consistently identified as relevant for both class A enzymes. Despite the modest statistical performance, the findings indicate that potent inhibitors tend to interact with essential catalytic residues, providing mechanistic insights. These interaction signatures will guide pharmacophore modeling and ZINC-based virtual screening, followed by in vitro validation to identify β-lactamase inhibitors as antibiotic adjuvants against AMR.