Abstract 974: AI-driven- structure-based drug discovery and validation of novel compounds as CTLA-4/B7 inhibitors

Abstract In recent years, CTLA-4/B7 immune checkpoints have been well known to suppress T-cell activation, with implications for patient survival under targeted therapeutics. Monoclonal antibodies targeting this immune checkpoint pathway have proven effective, and the use of recently FDA-approved antibodies clearly demonstrates their clinical and therapeutic value, showing promise across solid and hematologic cancers. However, the potential of small molecules derived from nature as future CTLA-4/B7 inhibitors remains largely unexplored. Naturally occurring compounds, as sources of chemical diversity, could yield novel leads for immune checkpoint inhibition. This approach may serve as one of the core strategies in modern immuno-oncology, complementing antibody-based therapies to improve patient outcomes in clinical settings. In this study, we developed CTLA-4/B7-specific machine learning (ML) and artificial intelligence (AI) models, along with 3D-QSAR models, to assist in ligand discovery as CTLA-4/B7 inhibitors. Structure-based virtual screening of compounds arising from natural sources was performed. Any compound predicted by our AI models as positive served as a candidate for subsequent structural biology studies. Protein-ligand interactions were investigated using molecular docking simulations, and the stability of resulting CTLA-4-inhibitor-B7 complexes was assessed using molecular dynamics simulations.To study biological relevance, we employed a PBMC-based functional assay and a tumor-relevant co-culture model (PBMC-B7 + tumor cells) to determine whether discovered leads could restore T-cell cytokine responses suppressed through CTLA-4/B7 engagement, complementing binding studies. Our ML/AI models prioritized flavones as new potential CTLA-4/B7 inhibitors, predicting binding affinity and favorable IC50 values. Molecular dynamics simulations further confirmed the stability of protein-flavone interactions. Experimentally, we selected chrysin, naringenin, and morin as representative molecules to validate that our workflow has potential to aid in the discovery of CTLA-4/B7 inhibitors. Functional assays revealed that flavones restored cytokine production suppressed as a result of CTLA-4/B7 engagement, thereby validating their immunological activity.Overall, our approach broadens therapeutic possibilities beyond antibody-based blockade, offering benefits in combination with existing therapeutics. The leads discovered may serve as starting points to develop derivatives that could advance next-generation immuno-oncology therapeutics to improve cancer care. Citation Format: Poonam Kalhotra, Tzayhri Gallardo-Velazquez, Guillermo Osorio-Revilla, Veera Chandra Sekhar Reddy Chittepu. AI-driven- structure-based drug discovery and validation of novel compounds as CTLA-4/B7 inhibitors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 974.