Rationalizing nose-to-brain drug delivery: Machine learning-guided optimization and mechanistic elucidation of olfactory deposition for nasal sprays.

Abstract

Nose-to-brain drug delivery via nasal sprays is severely limited by low olfactory deposition. In this study, we integrate machine learning with high-throughput experimentation to systematically optimize olfactory deposition of solid lipid nanoparticle-loaded nasal sprays (SLN-NS). By engineering the dispersed medium properties using five excipients at varying concentrations across 40 formulations, we establish a three-level interlocking correlation among dispersed medium properties, spray performance, and olfactory deposition. We demonstrate that basic physicochemical properties predominantly influence microscopic droplet characteristics, while dynamic rheological parameters dictate macroscopic spray morphology. A random forest model accurately predicted olfactory deposition and guided the design of an optimized formulation, achieving 19.77% olfactory deposition, representing a 4 to 20-fold improvement over conventional nasal sprays. In vivo studies confirmed enhanced brain delivery and superior therapeutic efficacy in animal models of migraine, Alzheimer's, and Parkinson's diseases. This work provides a rational framework for optimizing nasal spray formulations to overcome anatomical barriers for improved CNS drug delivery.