Latent structure modeling underlying theophylline tablet formulations using a Bayesian network based on a...

Area of Science:

• Pharmaceutical Science
• Computational Chemistry
• Data Science

Background:

• Quality by Design (QbD) in pharmaceutical development necessitates a scientific basis and defined design space.
• Understanding latent structures between formulation factors and drug product performance is crucial.

Purpose of the Study:

• To integrate thin-plate spline (TPS) interpolation, Kohonen's self-organizing map (SOM), and Bayesian networks (BN) for visualizing causal and latent factors in pharmaceutical responses.
• To quantitatively predict latent and response variables in theophylline tablet formulation.

Main Methods:

• Utilized TPS for nonlinear prediction of latent variables (compressibility, cohesion, dispersibility) and response variables (tensile strength, disintegration time).
• Employed SOM for clustering large datasets generated from pretableting blends and theophylline tablets.
• Applied BN to SOM clustering results to visualize the latent structure between causal factors, latent variables, and responses.

Main Results:

• Achieved high accuracy in predicting experimental values for latent and response variables.
• Successfully classified tablet data into distinct clusters using SOM.
• Demonstrated that BN analysis on SOM results effectively explains the underlying latent structure, confirming relationships between factors and responses.

Conclusions:

• The integrated TPS, SOM, and BN approach provides a robust method for understanding complex relationships in pharmaceutical formulation.
• This technique enhances the scientific rationale and design space establishment required by Quality by Design principles.
• Offers a deeper insight into the connections between formulation variables and the final drug product's performance.