Iterative hybrid model based optimization of rAAV production.

Area of Science:

• Biotechnology
• Gene Therapy Manufacturing
• Process Development

Background:

• Recombinant adeno-associated virus (rAAV) gene therapies require optimized HEK293 cell culture transfection for consistent production.
• Adapting transfection conditions is crucial for new rAAV serotypes or genetic payloads.

Purpose of the Study:

• To develop and validate an iterative, model-based experimental design approach for optimizing rAAV transfection conditions.
• To enhance rAAV production yields through improved transfection strategies.

Main Methods:

• An iterative model-based experimental design was employed, starting with shaker flask data to guide Ambr250 scale runs.
• Hybrid and historical models were developed using Ambr250 data to predict rAAV titer.
• A Bayesian Optimization framework combined these models for subsequent experimental design rounds.
• Transfer learning was applied across projects to enhance model predictive power.

Main Results:

• The model-based approach consistently yielded higher rAAV titers than benchmark methods (e.g., 32.4% increase in Project D for ELISA).
• Optimized transfection mixtures were achieved, leading to more efficient rAAV production.
• The iterative design process, incorporating increasing data and transfer learning, improved predictive accuracy.

Conclusions:

• The iterative model-based experimental design is an effective strategy for optimizing rAAV transfection and enhancing production yields.
• This approach offers a significant improvement over traditional statistical Design of Experiment methods.
• The integration of hybrid models, historical data, and transfer learning streamlines process development for gene therapies.