Advancing Direct Tablet Compression with AI: A multi-task framework for quality control, batch acceptance, and causal analysis

Area of Science:

• Pharmaceutical Manufacturing
• Drug Development
• Artificial Intelligence in Pharma

Background:

• The rise of Pharma 4.0 necessitates advanced AI solutions for pharmaceutical manufacturing.
• Direct Tablet Compression (DTC) is a critical technique reliant on precise control of physicochemical properties.
• Optimizing tablet quality requires understanding API, excipient, and process parameter interactions.

Purpose of the Study:

• To develop a novel multi-task AI framework for predicting tablet properties and batch acceptance in DTC.
• To provide insights for optimizing interactions to enhance tablet quality and manufacturing efficiency.
• To enable real-time monitoring, quality control, and regulatory compliance in pharmaceutical production.

Main Methods:

• A multi-task framework integrating regression, classification, and text generation.
• Application of statistical methods, Neural Networks (NN), Natural Language Processing (NLP), and generative AI (GenAI).
• Validation using the Harvard Dataverse V1 dataset of Fast Disintegrating Tablets (FDTs).

Main Results:

• Achieved 91.8% R-squared for regression tasks (predicting friability, hardness, disintegration time, water absorption ratio).
• Attained 95.5% accuracy for classification tasks (determining batch acceptance).
• Demonstrated superior performance compared to state-of-the-art methods.

Conclusions:

• The AI framework effectively predicts tablet properties and batch acceptance, improving DTC processes.
• The study provides a robust tool for real-time monitoring, quality control, and regulatory compliance.
• The developed framework aids in understanding and resolving manufacturing deviations for enhanced tablet quality.