摘要
Neurodegenerative diseases, including Huntington’s disease, Alzheimer’s disease, and Parkinson’s disease, are among the most challenging disorders to treat due to the restrictive nature of the blood-brain barrier (BBB), which blocks over 98% of potential therapeutics. Small interfering RNA (siRNA) presents a promising gene-silencing strategy, but clinical translation is hindered by degradation, off-target effects, and poor BBB permeability.This study presents SiExo, an AI-driven siRNA-loaded exosome platform engineered for high-efficiency CNS delivery. By integrating machine learning, molecular dynamics simulations, and Bayesian optimization, SiExo refines exosome lipid composition, ligand density, and siRNA stability for optimal therapeutic effect. Dual-ligand functionalization with transferrin and rabies virus glycoprotein enhances receptor-mediated BBB transcytosis, achieving a 3.5-fold increase in BBB penetration over unmodified exosomes. Functional testing in hCMEC/D3-based in vitro BBB models confirmed a 32% penetration rate (compared to 12% in controls, p < 0.01).AI-driven exosome design outperforms conventional lipid nanoparticles and viral vectors, yielding 2.8-fold greater siRNA retention and 1.9-fold higher neuron-specific uptake. In Huntington’s disease models, SiExo-mediated siRNA delivery resulted in a 62% reduction in mutant HTT mRNA expression, demonstrating significant gene-silencing potential. Bayesian optimization identified optimal exosome configurations with 92% predictive accuracy, accelerating non-immunogenic, patient-specific exosome engineering.This study establishes SiExo as a scalable CNS gene therapy platform, capable of supporting CRISPR/Cas9-based interventions for neurodegenerative and neuroinflammatory disorders. Future work will focus on in vivo validation and preclinical translation using patient-derived neural organoids and primate models, bridging the gap toward next-generation precision medicine for CNS disorders.