Work package 2

Project Title: Data streaming and augmentation for bioprocess development (WP2)

ESR 10

Host: Fujifilm Diosynth Biotechnologies, UK

Objectives: To develop a series of MVDA methods and related tools to exploit data from various recombinant yeast fermentation processes generated both at small (2-15 L) and commercial scale (1000-5000 L) which will improve process understanding and commercial scale process performance predictability thereby reducing reliance on a large number of commercial scale batches for process validation. This project aims to improve process understanding and predictability, at commercial scale, by exploring the wealth of data from small scale studies using appropriate tools and techniques and through comparison with large scale performance. The development data from FUJI at various scales will be initially used to develop the methodology. This will then be tested both on the commercial scale process data and also on industrial case studies from SAN and CHN.

Expected Results: Methodology for data augmentation for limited number of experiments (and their high variability due to exploring the design space) at small scale. Validation of the methodology on industrial commercial scale data. The methodology will feed into the final modelling framework.

 

 

Project Title: Early bioprocess development using PAT approaches (WP2)

ESR 2    

Host: Newcastle University, UK

Objectives: Reduce the time required to specify process flowsheet and operating conditions by utilising PAT modelling approaches and fundamental biological/biochemical knowledge of the production system/bioactive molecule. This is particularly important in new process development where more directed experimentation with process unit operations can significantly reduce the development times. Product characteristics (e.g. protein structure and physical properties) will be used as inputs to predict likely aggregation, solubility of purification issues that will indicate the most appropriate processing route. A number of model products will be used to develop such a toolbox, which will then be tested on the industrial case studies from FUJI and SAN.

Expected Results: PAT based methodology for process route prediction based on product characteristics. Validation of the methodology on industrial commercial scale data. The methodology will feed into the final modelling framework.

 

 

Project Title: Bioprocess risk assessment using a mechanistic modelling framework (WP2)

ESR 8    

Host: DTU, Denmark

Objectives: This project aims to develop and validate a risk assessment method using a mechanistic modelling approach. In particular this project will account for uncertainty and non-ideal sensor behaviour during processing. This will comply with the aims of PAT. First, the method will be demonstrated for recombinant yeast fermentations (proof of concept): variation in the most important model parameters will be evaluated on the basis of a series of experiments. This variation will then be implemented in the design, for example by using Monte Carlo simulation. Likewise, detailed characterization of all sensors and actuators in a lab-scale yeast fermentation will be performed, and non-ideal behaviour will be incorporated in the model library. Finally, a new design will be performed under uncertainty, resulting in a distribution of the outputs (e.g. product concentration). Such a distribution will then be interpreted to calculate the risk of not obtaining the required product quality.

Expected Results: Risk based methodology based on mechanistic, first principles modelling for evaluation of product quality performance. Validation of the methodology on industrial case studies (FUJI, SAN and CHN). The methodology will feed into the final modelling framework.