A Modular Mechanistic In Silico Model for In Vitro Transcription Process Yield and Product Quality Prediction.

Area of Science:

• Biotechnology
• Biochemical Engineering
• Computational Biology

Background:

• In vitro transcription (IVT) is vital for mRNA vaccine and therapeutic manufacturing.
• Optimizing mRNA yield and quality (capping, integrity) is mechanistically complex.
• Existing models lack the predictive power for complex IVT dynamics.

Purpose of the Study:

• To develop a modular mechanistic model of the IVT process.
• To improve scientific understanding and predictive capabilities of IVT.
• To support rational design and optimization of mRNA manufacturing.

Main Methods:

• Decomposition of IVT into six interconnected modules (initiation, elongation, termination, degradation, precipitation, pyrophosphate degradation).
• Development of kinetic models for each module based on biochemical principles and experimental data.
• Assembly of modules to capture coupled IVT dynamics, refined using machine learning (multivariate residual analysis, Shapley values) and Bayesian optimization.

Main Results:

• A scalable hybrid (mechanistic + machine-learning) modeling platform was established.
• Key IVT mechanisms were identified through sensitivity analysis.
• Efficient parameter estimation and model calibration were achieved using Gaussian-process-based Bayesian optimization.
• The model supports in silico experimentation and root-cause analysis.

Conclusions:

• The developed hybrid model provides a robust framework for understanding and optimizing IVT processes.
• This approach integrates heterogeneous data and accelerates model calibration.
• It facilitates the rational design and efficient manufacturing of mRNA-based products.