(660f) Mechanistic Process Modeling of in Vitro Transcription to Enable RNA Production Platforms

RNA-based biologicals are emerging as one of the most strategic technologies in the pharmaceutical industry. Recently, the COVID-19 pandemic brought them to the forefront of biomedical innovations, with Moderna’s and Pfizer-BIONTECH’s vaccines against SARS-COV-2 being the first RNA-based products to be approved for human use. It also highlighted the critical role that manufacturing technology plays in epidemic preparedness and disease management. The clinical success of these vaccines further delivers a hopeful message for accelerating the development of other RNA vaccines and therapeutics. In addition, these new products hold the potential for modular, disease-agnostic manufacturing platforms. The same raw materials, unit procedures, and analytical methods can be used for multiple targets, with only the sequence of the initial DNA construct to be modified. However, RNA heterogeneities require rethinking part of In Vitro Transcription (IvT) manufacturing strategy in a product-specific manner. RNA size, sequence features and secondary structures, as well as the type of nucleotides, need to be considered to optimise product quality and process capability. Additionally, the route of administration, targeted organ and dosing regimen can change the quality target product profiles (QTPPs), leading to further refinements of the optimal operating regions. Therefore, deeper process and product understanding appears essential to unlock the full platform potential of this technology. This study showcases how mechanistic process modelling can enhance this knowledge and underpin the development of modular, multiproduct RNA production processes.

Methodology

The principles of Quality by Digital Design (QbDD) have been applied holistically in this study. Initial model development is performed on a collated IvT database, encompassing numerous RNA constructs and broad process conditions. The modelling of key critical quality attributes (CQAs) has been prioritized based on up-to-date risk assessments. Overall, a differential-algebraic system of equations is implemented in PyoMo (Python 3.10) and combines kinetic and equilibrium reactions under partial-equilibrium approximations. This allows the predictions of multiple CQAs over time. Additionally, sensitivity analyses quantify product-process interactions. Then, model-based design-of-experiments (MB-DoE) can be easily derived to discriminate model structures and estimate product and platform-specific kinetic parameters.

Results

This work offers the first comprehensive modelling framework for multiproduct IvT. The integration of the different reaction inputs, including buffer components, allow us to combine different processes in the calibration dataset. RNA transcription and degradation are predicted based on the initial concentrations of plasmid DNA, free magnesium ions, NTPs, spermidine, DTT, pyrophosphatase, RNAse inhibitor and T7 RNA polymerase. Adding magnesium-driven precipitations is also essential to fully capture process variabilities. Based on this, a model-based optimization strategy is presented. Next, the impact of product-specific features on kinetic parameter estimates is explored in more details. For instance, the process conditions for more fragile, self-amplifying RNA systems are discussed. Then, in addition to conventional cross validation methods, calibration using IvT batch data accurately predicts fed-batch experiments and confirms model extrapolation capability. Multidimensional,iple, probabilistic design spaces are therefore defined. A workflow for product-specific parameter estimation is also introduced. Eventually, the potential use of these tools to support process and platform validation is reviewed.

Conclusions and discussion

Overall, no vaccines or biopharmaceutical have been developed and approved under such ambitious QbDD approach. This study confirms the platform potential of this new technology and provides deeper insights into the relationships between RNA product and processes. In the future, this holistic and mechanistic modelling approach is likely to be a canonical requirement for RNA platforms preapproval. Under this new paradigm, platform data and knowledge could be used extensively and reliably to accelerate development and regulatory submission processes, without compromisingwhile guaranteeing on product quality. This could be a major step towards the deployment of versatile RNA facilities and, ultimately, could reshape the future RNA manufacturing and supply landscape. To achieve this goal, the next step would be model integration into an end-to-end manufacturing process, potentially using process analytical technology (PAT) tools. Additionally, the incorporation of critical material attributes (CMAs) and RNA sequence through data-driven approaches could support broader knowledge and technology transfer.