UCL Discovery
UCL home » Library Services » Electronic resources » UCL Discovery

Development of predictive, experimentally driven mathematical models for CAR T cell therapy production

Glyn, Veronica Audrey Marie; (2023) Development of predictive, experimentally driven mathematical models for CAR T cell therapy production. Doctoral thesis (Ph.D), UCL (University College London).

[thumbnail of Glyn_10179521_Thesis.pdf] Text
Access restricted to UCL open access staff until 1 November 2025.

Download (30MB)


Since 2017 6 autologous CAR T therapies have been authorised by the FDA and EMA. Although these novel immunotherapies have achieved promising clinical results, the manufacturing process for CAR T therapies is not yet optimal, with up to 13% batch failures observed in clinical trials and high production costs, reflected by the treatment prices (>$375k). Manufacture typically follows a fixed recipe which does not account for patient-to-patient variability. Therefore, it is important to understand how the variable patient starting material (SM) impacts processing outcome, thereby providing the foundational knowledge for the design of robust adaptable process control strategies capable of compensating for SM variability. In this thesis, data-driven and knowledge-driven modelling techniques were used to enhance process and product understanding by finding correlations between starting materials, critical process parameters (CPPs), and critical quality attributes (CQAs). Historical data mining, correlation screening and principal component analysis (PCA), were used to identify CQAs based on their impact on CAR T cell function, after which they were prioritised based on existing process capability. The CQAs deemed of highest ranking were transduction efficiency (TE), CAR T cell number and the percentage of effector memory T cells re-expressing RA (TEMRA) at harvest. The historical dataset was further purposed using MVDA to identify process parameters with potential impacts on the aforementioned CQAs (potential critical process parameters, pCPPs). Subsequently, a DOE methodology was used to confirm the criticality of pCPPs and to quantify their relationships with CAR T cell number, TE, TEMRA and vector copy number (VCN) at harvest. Next, a Monod-based model of cell expansion and metabolism was developed to predict time-series profiles of viable cell, glucose, glutamine, lactate and ammonia concentrations. The model’s predictions were compared with experimental results generated by 5 donors. It successfully simulated average behaviour of CAR T cell expansion and metabolism, however, it could not simulate donor-specific behaviour. Of 18 model parameters, 4 were primarily associated with donor variability. Once adjusted, the model could describe the growth and metabolic profiles of all 5 donors. These parameters were: the yield of cells from glucose (Yx,glc), the yield of cells from glutamine (Yx,gln), the growth lag time (Ulag) and the critical lactate constant (CritLac). Moreover, the model was not appropriate for prediction of other product attributes, such as TE and TEMRA. To address the model’s limitations, multiple linear regression (MLR) modelling was used to predict TE, T cell number and TEMRA at harvest. This method, however, only allowed prediction of CQA at one harvest time point (assuming a constant production duration of 7 days) and did not take account of nutrient and metabolites. This limited the model’s extrapolative potential and use in feeding regime optimisation and control. Therefore, the lack of donor-specificity associated with the Monod-based expansion model was addressed with a hybrid modelling approach. Empirical functions were developed to predict the donor-specific parameters Yx,glc, Yx,gln, CritLac. The Monod-based expansion model was then updated with these predictions. As a result, the proposed hybrid model, could provide dynamic donor-specific predictions of in vitro CAR T cell expansion (viable cells) and metabolism (glucose, glutamine, lactate and ammonia). Simulated results aligned with experimental data derived from manufacture of CAR T cells from 5 different donors in GREXT M 6 well plates; the data included TEs, expansion curves (viable cell number) and metabolic profiles (glucose, lactate, ammonia). Finally, the predictive capabilities of the hybrid model were tested with data generated from 5 additional, independent batches, which were also produced in GREXT M 6 well plates. The experimental results were varied and contrasting as they were produced with different starting materials (donor cells) and processing conditions (MOI, transduced cell concentration and vectors). For example, within the test dataset the TEs and final viable T cell numbers achieved ranged between 44% and 92% and 81x106 and 211x106 , respectively. Despite the variation in experimental results, the model was able to capture the variation, showing proof of concept for the modelling strategy. This thesis provides data-driven insights into the impact of patient-to-patient variability on production outcome. Moreover, it provides a methodology to develop predictive cell culture models for autologous CAR T cell manufacturing, paving the way for optimisation studies and model-based control strategy design.

Type: Thesis (Doctoral)
Qualification: Ph.D
Title: Development of predictive, experimentally driven mathematical models for CAR T cell therapy production
Language: English
Additional information: Copyright © The Author 2023. Original content in this thesis is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) Licence (https://creativecommons.org/licenses/by-nc/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms. Access may initially be restricted at the author’s request.
UCL classification: UCL
UCL > Provost and Vice Provost Offices > UCL BEAMS
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Biochemical Engineering
URI: https://discovery.ucl.ac.uk/id/eprint/10179521
Downloads since deposit
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item