Presentation Information
[4DSP-12]Mechanistic Insights into AAV2 UF⁄DF: Predicting Performance with Data Efficient Experimentation
○Takao Ito1 (1. Merck Ltd.(An Affiliate of Merck KGaA, Darmstadt, Germany) (Japan))
Keywords:
viral vector,modeling,ultrafiltration,diafiltration,tangential flow filtration
Efficient downstream processing of adeno-associated virus (AAV) vectors increasingly depends on ultrafiltration and diafiltration (UF/DF), yet development of high-yield processes remains limited by scarce material availability and insufficient mechanistic understanding of membrane retention behavior. In early stages, UF/DF conditions are often selected empirically, despite the strong influence of flux dependent concentration polarization and wall concentration (Cw) on viral retention and impurity reduction. To address this challenge, this study evaluates a mechanistic model based on stagnant film theory to improve process understanding and provide a more rational basis for AAV2 UF/DF development.
The model integrates mass-balance relationships for retentate volume, concentration, and permeate composition, together with a semi empirical description of flux–TMP behavior. Concentration polarization is incorporated through estimation of wall concentration Cw, enabling flux-dependent prediction of sieving and yield. These simplified mechanistic components are designed to maximize interpretability and relevance for early-phase development, where experimental throughput is limited. To conserve AAV material, bacteriophage was employed as a surrogate for preliminary calibration. The calibrated model was subsequently applied to AAV2 UF/DF data generated under constant permeate flow in three independent runs.
Simulations reproduced TMP progression with high fidelity across the full UF/DF sequence (overall RMSE 0.03195). After calibrating the intrinsic sieving parameter using the first AAV run, the model predicted AAV2 yields under different flux conditions, supporting the hypothesis that flux-driven changes in Cw dominate AAV retention behavior. This agreement confirms the utility of the stagnant film framework as a physically grounded explanation for variations in UF/DF performance.
Overall, the findings show that a simplified mechanistic model can describe the primary factors governing AAV2 UF/DF behavior while requiring minimal experimental resources. The approach offers a practical means to interpret process performance, guide selection of flux and membrane loading, and reduce reliance on empirical trial-and-error during early development. Although further refinement will be required for more comprehensive empirical demonstration, the combination of surrogate-feed calibration and mechanistic modeling provides a resource-efficient pathway for accelerating AAV UF/DF process development.
The model integrates mass-balance relationships for retentate volume, concentration, and permeate composition, together with a semi empirical description of flux–TMP behavior. Concentration polarization is incorporated through estimation of wall concentration Cw, enabling flux-dependent prediction of sieving and yield. These simplified mechanistic components are designed to maximize interpretability and relevance for early-phase development, where experimental throughput is limited. To conserve AAV material, bacteriophage was employed as a surrogate for preliminary calibration. The calibrated model was subsequently applied to AAV2 UF/DF data generated under constant permeate flow in three independent runs.
Simulations reproduced TMP progression with high fidelity across the full UF/DF sequence (overall RMSE 0.03195). After calibrating the intrinsic sieving parameter using the first AAV run, the model predicted AAV2 yields under different flux conditions, supporting the hypothesis that flux-driven changes in Cw dominate AAV retention behavior. This agreement confirms the utility of the stagnant film framework as a physically grounded explanation for variations in UF/DF performance.
Overall, the findings show that a simplified mechanistic model can describe the primary factors governing AAV2 UF/DF behavior while requiring minimal experimental resources. The approach offers a practical means to interpret process performance, guide selection of flux and membrane loading, and reduce reliance on empirical trial-and-error during early development. Although further refinement will be required for more comprehensive empirical demonstration, the combination of surrogate-feed calibration and mechanistic modeling provides a resource-efficient pathway for accelerating AAV UF/DF process development.
Comment
To browse or post comments, you must log in.Log in