Data-smart machine learning methods for predicting composition-dependent Young's modulus of pharmaceutical compacts | JoVE Visualize

Area of Science:

Pharmaceutical Sciences
Materials Science
Computational Chemistry

Background:

Predicting mechanical properties of Active Pharmaceutical Ingredient (API) and excipient blends is crucial for efficient formulation design.
Traditional methods involve extensive trial-and-error, increasing development time and cost.
Machine learning (ML) offers a data-driven approach to accelerate model development, provided sufficient historical data is available.

Purpose of the Study:

To evaluate the efficacy of linear and nonlinear machine learning models in predicting the Young's modulus (YM) of directly compressed pharmaceutical powder blends.
To determine if YM can be accurately predicted using only the true density of the API, alongside known excipient properties.
To demonstrate the application of predictive models in facilitating Quality-by-Design (QbD) principles for pharmaceutical formulations.

Main Methods:

Developed and trained linear and nonlinear ML models using data from compacts of three BCS Class I APIs and two excipients.
Varied blend compositions by adjusting drug loadings and excipient ratios, and compacted to five nominal solid fractions.
Assessed model prediction accuracy using three distinct cross-validation schemes.

Main Results:

Linear and nonlinear ML models demonstrated significant utility in predicting the Young's modulus of the tested powder blends.
The models achieved accurate predictions based solely on the true density of the API and known excipient characteristics.
Cross-validation confirmed the robustness and reliability of the predictive models across various blend compositions and compaction states.

Conclusions:

Machine learning models can effectively predict the Young's modulus of pharmaceutical powder blends, reducing the need for extensive experimental screening.
The true density of the API is a key parameter for accurate mechanical property prediction in directly compressed formulations.
This predictive capability supports the implementation of Quality-by-Design in pharmaceutical formulation development, enhancing efficiency and control.