CHO cell factory modelling

工艺开发通常基于为单抗生产开发的现有平台工艺，这些工艺在早期选择高产克隆的基础上提供了高滴度的起点。进一步的工艺优化导致通常预先设计的操作(主要是反应器进料)策略适用于所讨论的细胞系/产品。优化过程通常是经验性的，相关能力存在于组织中的几个人手中。在学术界和越来越多的工业中，数学模型被用来快速探索不同的运营策略，并加快流程优化。结构化和以知识为基础的建模战略将可以在不同的细胞系和工艺条件之间转移，以便能够在生物工艺开发和运行期间作出充分合理的决策。一个例子是伦敦帝国理工学院和MedImmune最近提出的细胞生长、单抗产量和糖基化的动力学模型(Kotidis等人，2019年，生物技术和生物工程，116(7)：1612-1626)。该模型被用来通过计算筛选超过8,000组工艺条件，以识别满足两个限制(最小滴度和半乳糖化程度)的子空间，并通过实验验证了发现(Kotidis等人，2019，Computers&Chemical Engineering，125：558-568)。虽然模型成功地预测了是否满足约束条件，但我们发现它不能完全定量地预测。这在某种程度上并不令人惊讶，因为该模型，像所有的动力学模型一样，在其结构和参数值上是固定的，因为我们知道产生细胞的代谢行为在整个培养期间并不是恒定的。我们的目标是建立一个模块化的细胞工厂建模平台，它将动力学模型与通量平衡分析相结合，以定量和准确地描述(I)细胞培养过程中细胞外和细胞内代谢产物之间的化学计量关系是如何变化的，(Ii)这些化学计量比的变化如何影响CHO细胞生长的动力学，以及(Iii)细胞代谢、产品生物合成、组装和分泌如何结合在一起产生不同的生产力水平和单抗产品糖基化。

Process development is typically based on existing platform processes developed for mAb production, which give a starting point for high titres based on an earlier selection of high-producing clones. Further process optimisation results in often pre-designed operational (primarily reactor feeding) strategies applicable to the cell line/product in question. The optimisation process is typically empirical and the associated capability lies within a few individuals in an organisation. In academia and increasingly in industry, mathematical models are used to rapidly explore different operating strategies and expedite process optimisation. Structured and knowledge-based modelling strategies would be transferrable across different cell lines and process conditions to enable fully rational decision-making during bioprocess development and operation. An example is the recently proposed kinetic model of cell growth, mAb productivity and glycosylation developed by Imperial College London and MedImmune (Kotidis et al., 2019, Biotechnology and Bioengineering, 116(7):1612-1626). The model was used to screen over 8,000 sets of process conditions computationally to identify the subspace the met two constraints (minimum titre and extent of galactosylation) and the findings were verified experimentally (Kotidis et al., 2019, Computers & Chemical Engineering, 125: 558-568). Although the model successfully predicted whether the constraints were met or not, we found that it is not fully quantitatively predictive. This is somewhat unsurprising because the model, like all kinetic models, is fixed in its structure and parameter values, when it is known that the metabolic behaviour of producer cells is not constant throughout the culture period.Our goal is to generate a modular cell factory modelling platform that integrates kinetic models with flux balance analysis to quantitatively and accurately describe (i) how the stoichiometric relationships between extracellular and intracellular metabolites vary during cell culture processes, (ii) how these shifts in stoichiometry influence the dynamics of CHO cell growth, and (iii) how cellular metabolism, product biosynthesis, assembly and secretion, all combine to yield different productivity levels and mAb product glycosylation.