Computational Modeling of Chimeric Antigen Receptor–Natural Killer Cell Therapy Targeting B7 Homolog 3 (B7-H3) Receptor in Neuroblastoma

Neuroblastoma, a pediatric malignancy originating from neural crest cells, remains one of the leading causes of cancer-related mortality in children. Current treatment approaches—surgery, chemotherapy, and radiotherapy—are limited by systemic toxicity and lack of tumor specificity. In this study, we employed computational modeling to design and evaluate a chimeric antigen receptor–natural killer (CAR-NK) cell therapy targeting the B7-H3 (V-set domain-containing T-cell activation inhibitor 1) receptor, an immune checkpoint protein overexpressed in neuroblastoma. The amino acid sequence of B7-H3 (UniProt ID: Q5ZPR3) was retrieved and modeled using AlphaFold 3, revealing a well-defined extracellular domain suitable for receptor binding. GRAMM-based docking simulations demonstrated a strong and stable interaction between the single-chain variable fragment (scFv) of the CAR construct and the B7-H3 recep- tor, driven by hydrogen bonding and hydrophobic interactions among residues Tyr, Phe, Lys, and Asp. Binding-site predictions from PrankWeb identified residues 215–260 as the primary ligand-accessible re- gion, consistent with the docking interface, while Protter visualization confirmed the transmembrane topology of B7-H3.  The in silico results suggest that CAR-NK binding to B7-H3 can effectively trig-   ger perforin- and granzyme-mediated apoptosis, enhancing tumor-specific cytotoxicity with minimal off- target effects. This study establishes a reproducible computational workflow integrating AI-based struc- ture prediction and docking analysis to guide rational CAR design. The findings provide a foundational framework for developing B7-H3–directed CAR-NK immunotherapy as a safe and potent treatment strat- egy for neuroblastoma.