A novel preformulation tool to group microcrystalline celluloses using artificial neural network and data clustering

Area of Science:

• Materials Science
• Chemical Engineering
• Data Science

Background:

• Microcrystalline celluloses (MCCs) are widely used excipients in pharmaceutical formulations.
• Understanding MCC properties is crucial for optimizing drug delivery and formulation stability.
• Current methods for MCC characterization and grouping can be subjective.

Purpose of the Study:

• To develop an objective method for grouping microcrystalline celluloses (MCCs).
• To utilize artificial neural networks (ANN) and data clustering for MCC classification.
• To identify key physical properties influencing MCC-water interactions.

Main Methods:

• Employing a radial basis function (RBF) network to model MCC torque measurements from mixer torque rheometry (MTR).
• Applying discrete incremental clustering (DIC) to group MCCs based on RBF network outputs.
• Analyzing the correlation between torque parameters (Torque(max)) and MCC physical properties.

Main Results:

• Successfully grouped 11 MCCs into 2 or 3 distinct clusters based on rheological data.
• Identified bulk and tapped densities as critical factors in water-MCC interactions.
• Observed that denser MCCs (e.g., Avicel PH 301, PH 302) exhibit higher sensitivity to water content.

Conclusions:

• A combined ANN and DIC approach provides an objective method for MCC grouping.
• This methodology facilitates the preliminary assessment of novel or uncharacterized MCCs.
• The study reveals key properties governing MCC performance in aqueous environments.