Targeted probiotic tabletting: A hybrid active learning and finite element modelling approach for process optimisation

Bide Wang ¹, Xilu Wang ², Oleksiy V Klymenko ¹

⁰School of Chemistry and Chemical Engineering, University of Surrey, Guildford GU2 7XH, UK.

International Journal of Pharmaceutics: X +

|

November 10, 2025

View abstract on PubMed

Summary

Optimizing probiotic tablet production is challenging. A new method using active learning and finite element modeling rapidly identifies ideal compression settings to maximize probiotic survival during tabletting.

Area Of Science

Pharmaceutical Sciences
Computational Modeling
Biotechnology

Background

Tablets are an effective delivery method for probiotics.
Previous research identified compression pressure, speed, and precompression as key factors for probiotic survival.
Experimental studies are time-consuming, limiting optimization of individual parameters.

Purpose Of The Study

To develop a systematic approach for identifying optimal process parameters for probiotic survival during tabletting.
To overcome the limitations of traditional experimental methods in pharmaceutical formulation.
To accelerate the optimization of probiotic tabletting processes.

Main Methods

Integrated active learning (AL) with Gaussian process regression (GPR) and finite element (FE) modeling.
Utilized an FE model to generate data for predicting probiotic viability during tabletting.
Employed global random sampling and threshold filtering to identify optimal parameter regions.

Main Results

Achieved high prediction performance (R²=0.96) for probiotic survival rate after 78 iterations.
Successfully identified regions for near-optimal probiotic survival.
Generated survival rate maps revealing interplay between survival and tablet mechanical performance.

Conclusions

Hybrid data-driven and first-principles modeling offers a robust strategy for optimizing probiotic tabletting.
This approach accelerates pharmaceutical development by enabling efficient process optimization.
The study highlights the potential of computational methods in enhancing drug delivery system design.

Keywords:

Active learning Finite element method Gaussian process regression Probiotic Tabletting