In bioprocess engineering the Qualtiy by Design (QbD) initiative encourages the use of models to define design spaces. However, clear guidelines on how models for QbD are validated are still missing. In this review we provide a comprehensive overview of the validation methods, mathematical approaches, and metrics currently applied in bioprocess modeling. The methods cover analytics for data used for modeling, model training and selection, measures for predictiveness, and model uncertainties. We point out the general issues in model validation and calibration for different types of models and put this into the context of existing health authority recommendations. This review provides a starting point for developing a guide for model validation approaches. There is no one-fits-all approach, but this review should help to identify the best fitting validation method, or combination of methods, for the specific task and the type of bioprocess model that is being developed.

Maximizing the value of each available data point in bioprocess development is essential in order to reduce the time-to-market, lower the number of expensive wet-lab experiments, and maximize process understanding. Advanced in silico methods are increasingly being investigated to accomplish these goals. Within this contribution, we propose a novel integrated process model procedure to maximize the use of development data to optimize the Stage 1 process validation work flow. We generate an integrated process model based on available data and apply two innovative Monte Carlo simulation-based parameter sensitivity analysis linearization techniques to automate two quality by design activities: determining risk assessment severity rankings and establishing preliminary control strategies for critical process parameters. These procedures are assessed in a case study for proof of concept on a candidate monoclonal antibody bioprocess after process development, but prior to process characterization. The evaluation was successful in returning results that were used to support Stage I process validation milestones and demonstrated the potential to reduce the investigated parameters by up to 24% in process characterization, while simultaneously setting up a strategy for iterative updates of risk assessments and process controls throughout the process life-cycle to ensure a robust and efficient drug supply.

A robust understanding of structure-function relationships (SFRs) is fundamental for tailored and robust chemistry, manufacturing, and control (CMC) activities. Yet, conventional forced degradation studies (FDS) often employ a one-factor-at-a-time (OFAT) methodology, which commonly leads to correlated modifications and therefore limited reliability of SFR determination missing statistical certainty. By applying a forced-degradation-study design of experiments (FDS-DoE) framework to trastuzumab, methodological advantages are underscored, including improved data interpretation and consistency with literature findings, while adding deeper insights to the antibodies SFRs. As the choice of stress conditions in the design of FDS studies is critically important to establish meaningful SFRs, we present a detailed experimental comparison of harsh vs. moderate CMC development relevant stress conditions applying the FDS-DoE workflow. While harsh conditions mainly reflect typical FDS protocols and test the boundaries of a molecule's resilience, the moderate conditions were designed to simulate CMC-relevant environments better. Furthermore, we introduce a critical quality attribute (CQA) assessment framework that uses partial least square regression to systematically categorize modifications according to their statistical significance (via permutation testing) and predicted potency loss. The combination of multifactorial FDS-DoE with this CQA assessment framework increased statistical power over traditional OFAT methods, establishing a predictive, automated and scalable strategy for identifying relevant SFRs and CQAs, a capability especially advantageous for macromolecules in biopharmaceutical development.

Intra-experimental factor setting shifts in intensified design of experiments (iDoE) enhance understanding of bioproduction processes by capturing their dynamics and are thus essential to fulfill quality by design (QbD) ambitions. Determining the influence of process history on the cellular responses, often referred to as memory effect, is fundamental for accurate predictions. However, the current iDoE designs do not explicitly consider nor quantify the influence of process history. Therefore, we propose the one-factor-multiple-columns (OFMC)-format for iDoE planning. This format explicitly describes stage-dependent factor effects and potential memory effects as across-stage interactions (ASIs) during a bioprocess. To illustrate its utility, an OFMC-iDoE that considers the characteristic growth phases during a fed-batch process was planned. Data were analyzed using ordinary least squares (OLS) regression as previously described via stage-wise analysis of the time series and compared to direct modeling of end-of-process outcomes enabled by the OFMC-format. This article aims to provide the reader with a framework on how to plan and model iDoE data and highlights how the OFMC-format simplifies planning, and data acquisition, eases modeling and gives a straightforward quantification of potential memory effects. With the proposed OFMC-format, optimization of bioprocesses can leverage which factor settings are most beneficial in which state of the mammalian culture and thus elevate performance and quality to the next level.

Within process development, numerous experimental studies are undertaken to establish, optimize and characterize individual bioprocess unit operations. These studies pursue diverse objectives such as enhancing titer or minimizing impurities. Consequently, Design of Experiment (DoE) studies are planned and analyzed independently from each other, making it challenging to interlink individual data sets to form a comprehensive overview at the conclusion of the development process. This paper elucidates the methodology for constructing a Life-Cycle-DoE (LDoE), which integrates data-driven process knowledge through design augmentations. It delves into the strategy, highlights the challenges encountered and provides solutions for overcoming them. The LDoE approach facilitates the augmentation of an existing model with new experiments in a unified design. It allows for flexible design adaptations as per the requirements of subject matter experts (SME) during process development, concurrently enhancing model predictions by utilizing all available data. The LDoE boasts a broad application spectrum as it consolidates all data generated within bioprocess development into a single file and model. The study demonstrates that the LDoE approach enables a process characterization study (PCS) to be performed solely with development data. Furthermore, it identifies potentially critical process parameters (pCPPs) early, allowing for timely adaptations in process development to address these challenges.

Determining process variances in biopharmaceutical manufacturing is challenging due to limited data availability. To address this, we introduce a Bayesian hierarchical model designed for meta-analysis of process variance. This approach can improve process variance estimation by integrating data from multiple products, providing more reliable estimates of critical quality attributes in cases of data scarcity. Additionally, our model aids in evaluating process models, ensuring quality in process development. The paper demonstrates the new method using a simulation study, showcasing its potential to leverage historical data for both upstream and downstream phases of future CMC drug development. The new statistical model has great potential to expedite the market introduction of therapies while ensuring patient safety, allowing new treatments to reach patients more quickly without compromising quality or efficacy.