Machine learning strengthened formulation design of pharmaceutical suspensions

Area of Science:

• Pharmaceutical Sciences
• Materials Science
• Computational Chemistry

Background:

• Long-acting injectables (LAIs) are crucial for managing chronic conditions.
• Current LAI formulation development relies heavily on trial-and-error.
• Controlling particle size in nano- and microsuspensions is key for drug release.

Purpose of the Study:

• To systematically investigate formulation parameters influencing nano- and microsuspensions.
• To apply statistical and machine learning (ML) strategies for formulation optimization.
• To develop an explainable ML model for predicting formulation stability.

Main Methods:

• Full-factorial milling experiments were conducted.
• Statistical analysis identified significant formulation factors.
• A machine learning classification model was built and validated using 72 data points.
• Shapley additive explanations (SHAP) were used for model interpretation.

Main Results:

• Stabilizer concentration was a highly significant factor (p < 0.001) affecting median suspension diameter (D50).
• An ML model achieved high prediction accuracy (0.91) and F1-score (0.91) for formulation stability.
• SHAP analysis highlighted the critical roles of stabilizer concentration and milling bead size.

Conclusions:

• Explainable ML modeling can significantly enhance understanding of nano- and microsuspension design.
• This approach offers a data-driven alternative to traditional trial-and-error methods.
• The developed ML model aids in optimizing LAI formulations for improved drug delivery.