Application of AI in Tablet Development: An Integrated Machine Learning Framework for Pre-Formulation Property...

Area of Science:

• Pharmaceutical Sciences
• Computational Chemistry
• Materials Science

Background:

• Tablet development involves optimizing multiple quality attributes under budget constraints.
• Formulation-property relationships in mixture systems are highly nonlinear.
• Pre-formulation decision-making requires advanced predictive models.

Purpose of the Study:

• To propose an AI framework for organizing tablet formulation, process, and raw material data.
• To enrich conventional features with physically motivated mixture descriptors.
• To support pre-formulation decision-making by predicting tablet quality attributes.

Main Methods:

• Developed an AI framework integrating formulation, process, and material property data.
• Constructed mixture-level scalar descriptors and incorporated particle size distribution (PSD) summaries.
• Compared three feature sets (MP, MPD, MPDD) using six machine learning models and cross-validation strategies.

Main Results:

• Mixture-descriptor augmentation improved predictions for tablet hardness and disintegration time in interpolation settings.
• Smaller gains were observed for flow function, with mixed effects for cohesion and thickness.
• Extrapolation-oriented evaluation showed potential improvements for hardness but degradation for disintegration time prediction.

Conclusions:

• The proposed AI framework and feature enrichment strategies can aid pre-formulation decision-making.
• Careful selection and dimensionality control of descriptors are crucial for robust extrapolation.
• The study highlights the need for robust AI models in complex pharmaceutical mixture development.