Machine Learning Guided Optimization of an Oral Microemulsion System: A Bayesian Optimization Approach.

Area of Science:

• Pharmaceutical Sciences
• Drug Delivery Systems
• Formulation Development

Background:

• Oral microemulsions (MEs) enhance oral bioavailability of poorly water-soluble drugs.
• Traditional ME formulation is time-consuming and resource-intensive, often yielding suboptimal results.
• There is a need for efficient methods to streamline ME formulation design.

Purpose of the Study:

• To employ a batch Bayesian Optimization strategy for designing optimized microemulsion formulations.
• To accelerate the identification of stable and high-performing microemulsions.
• To improve the physicochemical properties and therapeutic performance of microemulsions.

Main Methods:

• A batch Bayesian Optimization strategy was implemented over five iterations (five experiments per batch).
• A two-phase modeling approach navigated a complex experimental space with multiple oils, surfactants, and cosurfactants.
• A training dataset of 22 experiments was utilized to build the optimization model.

Main Results:

• Five high-performing blank microemulsions were successfully designed.
• Four of the identified microemulsions demonstrated physical stability for up to 30 days.
• When loaded with model drugs, three microemulsions showed high drug loading, good stability, and improved in-vitro permeability.

Conclusions:

• Bayesian optimization significantly streamlines the design of microemulsion formulations.
• The developed method efficiently identifies stable and effective microemulsions for drug delivery.
• The optimized microemulsions show promise for the oral delivery of poorly soluble drugs.