Robust design of chemical processes with application to bio-manufacturing

引入越来越多的生物产品，如新的单克隆抗体和新的疫苗，显然需要新的强大和有效的生物工艺来生产这些产品。在哺乳动物细胞培养中生产的单克隆抗体（mab）目前是一种主要的生物制品，2012年全球销售额为450亿美元，预计年需求增长率为9.5%。例如，大约有28个单抗已被批准，例如西妥昔单抗（转移性癌症），曲妥珠单抗（乳腺癌），还有350多个处于临床试验的不同阶段。疫苗生产是另一个主要的战略领域，正在通过引进新疫苗而不断恢复活力。本提案研究了生物过程稳健设计的不同方面，并以之前资助期在两个领域进行的研究为基础：i)基于代谢通量平衡的模型和ii)使用不完美模型的化学过程的同时最优控制和设计。美国食品和药物管理局（FDA）最近采用了设计质量（QBD）方法，即整个生物过程是先验设计和认证的，以在称为设计空间的范围内提供最终产品的质量和数量。在QBD方法中，一个过程可以在设计空间内进行调整，以适应原材料的可变性和干扰，而传统方法中，过程大多是用固定的操作配方操作的。为了应用QBD方法，有必要寻找最佳设计和最佳操作条件，同时考虑到输入的先验预期可变性，例如原材料，以及用于优化和控制的模型的不准确性。本研究的长期目标是开发一种基于模型的新生物过程优化设计方法或现有生物过程的优化升级方法，这些方法对模型误差和过程变化具有鲁棒性。通过明确考虑对模型误差和过程可变性的鲁棒性，该工作非常适合QBD方法。为了实现我们的长期目标，稳健的建模、优化和控制工具将通过以下四个项目开发：开发哺乳动物细胞动态代谢通量平衡模型（DMFB）用于优化。2。容忍模型误差和过程变化的生物过程优化设计。3。DMFBs和多元统计模型的集成，用于生物过程的优化设计或升级。基于动态代谢通量模型的同步优化设计与非线性模型预测控制（NMPC）我建议在生物技术过程的建模、控制和优化的不同领域培养2名博士（在授予期结束时再增加2名）和2名硕士。这项工作包括动态代谢通量分析、跑对跑优化和非线性预测控制领域的理论发展，以及校准所提出的模型和验证优化程序的实验。将考虑应用于哺乳动物细胞培养生产单克隆抗体和百日咳疫苗生产工艺。总的来说，提出的基于模型的稳健设计工具将以两种方式使该行业受益：i-通过提出一种生物过程设计方法，确保按照QBD理念预先确定的生产力和质量水平；ii-通过培训HQP，使其具备将过程系统工程工具（建模、控制、统计和优化）应用于生物技术过程的多种专业知识。

The introduction of an increasing number of bio-products, such as new monoclonal antibodies and new vaccines, has created a clear need for new robust and efficient bioprocesses to manufacture these products. mAbs (monoclonal antibodies) produced in mammalian cell cultures are currently a predominant bioproduct with world-wide sales of $45 billion in 2012 and with a predicted annual growth in demand of 9.5%. For instance, approximately 28 mAbs have been approved, e.g cetuximab (metastatic cancers), trastuzumab (breast cancer), while 350 more are in various stages of clinical trials. Vaccine manufacturing is another major strategic area that is being continuously re-energized by the introduction of new vaccines. This proposal investigates different aspects of robust design of bioprocesses and builds upon research conducted in the previous funding period in two areas: i) metabolic flux balance-based models and ii) simultaneous optimal control and design of chemical processes using imperfect models.The US Food and Drug Administration (FDA) has recently adopted the quality-by-design (QBD) approach whereby the entire bioprocess is a priori designed and certified to provide quality and quantity of the final product within a range of bounds referred to as the design space. In the QBD approach a process can be adjusted within the design space to accommodate variability in raw materials and disturbances in contrast with the traditional approach where the process is operated with mostly fixed operating recipes. To apply the QBD approach it is necessary to search for an optimal design and optimal operating conditions while taking into account a priori expected variability in inputs, e.g. raw materials, and inaccuracies of the models to be used for optimization and control. The long term objective of this research is to develop a method for model-based optimal designs of new bioprocesses or optimal upgrades of existing ones that are robust to model errors and process variations. By explicitly considering robustness to model error and process variability, the work closely fit the QBD approach. To achieve our long term goal, robust modeling, optimization and control tools will be developed via the following four projects:I. development of dynamic metabolic flux balance models (DMFB) for mammalian cells to be used for optimization. II. optimal design of bioprocesses tolerant to model errors and process variabilities. III. integration of DMFBs and multivariate statistical models for optimal design or upgrade of bioprocesses. IV. simultaneous optimal design and nonlinear model predictive control (NMPC) based on dynamic metabolic flux models. I am proposing to train 2 PhD’s (and additional 2 towards the end of the granting period) and 2 MASc’s in different areas of modeling, control and optimization of biotechnological processes. The work includes both theoretical developments in the areas of dynamic metabolic flux analysis, run-to-run optimization and nonlinear predictive control as well as experimentation to calibrate the proposed models and to validate the optimization procedures. Applications to a mammalian cell culture producing mAb and a whooping cough vaccine manufacturing process will be considered.Overall, the proposed model-based robust design tools will benefit the industry in two ways: i-by proposing an approach to bioprocess design that ensures predefined levels of productivity and quality as per the QBD idea and ii- by training HQP with versatile expertise in the application of process systems engineering tools (modeling, control, statistics and optimization) to biotechnological processes.