Computational Exploration of Biphenyl Derivatives as PD‐L1 Inhibitors for TNBC Using Docking, Molecular Dynamics, MM‐GBSA, DFT, and AI/ML‐Based QSAR Modeling

Triple negative breast cancer (TNBC) remains a leading cause of cancer‐related mortality among women, creating an urgent need for new therapeutic strategies. This study aims to design potential PD‐L1 inhibitors using an in silico approach, integrating molecular docking with quantitative structure–activity relationship (QSAR) modeling in the discovery of promising candidates for immunotherapy in TNBC. Around 12 compounds were selected based on binding affinity equal to or greater than 9.5 kcal/mol. Binding analysis was done to select a lead compound. The reference ligand exhibited a binding affinity of −7.1 kcal/mol at the PD‐1/PD‐L1 interface, whereas the selected compound 3 with improved binding affinity (−13.4 kcal/mol) showed key interactions with critical residues Tyr56, Met115, Ala121, and Asp122, suggesting a comparatively stronger predicted binding affinity toward the PD‐L1 interface relative to the reference compound, within the limitations of docking‐based evaluation. The compound 3‐7DY7 complex demonstrated enhanced stability during 100 ns molecular dynamics (MD) simulations, maintaining low RMSD and strong hydrogen–bond interactions, supported by a favorable MM‐GBSA binding free energy (−118.994 kcal/mol). Compound 3 exhibited a slightly higher HOMO energy (−0.184 eV) compared to reference (−0.219 eV), indicating a greater ability to donate electrons. Consequently, compound 3 also showed a higher nucleophilicity index ( N = 8.935 eV), suggesting that it is more susceptible to electrophilic attack. Based on the QSAR analysis, the calculated pIC 50 value of compound 3 was 8.8. The QSAR modeling ( R 2 = 0.974, RMSE = 0.244, MAE = 0.167) confirmed robust predictive capacity. These results highlight compound 3 as a promising inhibitor of PDL‐1. This candidate could be prioritized for subsequent chemical synthesis and in vitro bioassays to confirm the computational findings.