AI-designed estrogen analogues encapsulated in pH-responsive β-cyclodextrin nanosponges for localized vaginal delivery in cervical cancer

Abstract

Systemic therapy for cervical cancer is frequently limited by off-target toxicity and inadequate localized drug exposure. We report an integrated platform combining generative artificial intelligence (AI) and nanotechnology to design receptor-selective estrogen analogues and deliver them via a vaginal, pH-responsive nanocarrier for localized hormone therapy. Methods: Generative deep learning models were used to design novel estrogenic ligands optimized for binding affinity and selectivity toward estrogen receptor subtypes (ERα and ERβ) implicated in cervical malignancies. Top candidates were evaluated by in silico docking and molecular dynamics to predict receptor interactions and stability. Selected analogues were synthesized and encapsulated in β-cyclodextrin nanosponges engineered for pH-triggered release in the vaginal microenvironment. Nanosponges were characterized for particle size, encapsulation efficiency, and release kinetics. In vitro assays on cervical cancer cell lines assessed cytotoxicity, receptor activation/inhibition, and cellular uptake; systemic absorption potential was evaluated in simulated mucosal models. Results: AI-optimized ligands showed improved predicted affinity and subtype selectivity versus reference estrogens. Nanosponges achieved high encapsulation efficiency and demonstrated sustained, pH-dependent release with approximately 70– 85% cumulative release over 72 hours under vaginal-like conditions. In vitro, the formulation produced enhanced cytotoxic and/or antiproliferative effects in ER-expressing cervical cancer cells with minimal markers of nonspecific toxicity. Simulated mucosal studies indicated limited systemic permeation, suggesting favorable localization. Conclusions: The combination of AI-driven molecular design and β-cyclodextrin nanosponge delivery enables targeted, sustained vaginal administration of estrogen analogues for cervical cancer, potentially maximizing tumor exposure while minimizing systemic side effects. Further in vivo evaluation and safety profiling are warranted to advance this precision oncology approach.