Raman analysis of in process bioreactors to develop enhanced bioprocess control strategies

除了监测关键氨基酸浓度外，哺乳动物培养物的代谢谱分析还可以更好地了解整个生物反应器中细胞的生理状态。这些信息对于定制氨基酸饲料的开发是无价的，可以针对单个细胞系进行定制，以提高生产力并最大限度地减少产品的异质性。然而，代谢谱分析的一个主要挑战是为大量细胞培养生成时间过程代谢谱的成本和挑战。该项目旨在通过将下一代高通量（HT）拉曼光谱设备的功能与先进的微生物反应器系统相结合来克服这一障碍。这种高温拉曼光谱装置只需要25-100 μ l的小体积，就能产生高分辨率的拉曼光谱。除了所有相关的生物反应器离线测量外，这些光谱还可以预测氨基酸浓度，包括葡萄糖，乳酸，细胞密度和产物浓度，这为开发更好的控制策略开辟了重要的机会。这项工作将生成一个数据集，用于训练和应用先进的多变量数据分析（MVDA）和机器学习（ML）模型，以建立拉曼光谱与所有可用的离线测量值之间的相关性。这些测量可以纳入控制策略，实现必需氨基酸浓度的在线控制，并为开发定制饲料提供基础，以提高生物反应器性能并保持所需的产品质量规格。本研究的主要目标之一是开发先进的算法，以自主生成鲁棒的数学模型，实现关键工艺参数的近实时预测，以支持先进的控制决策并增强制造操作。这些新颖的计算模型将能够在计算机上预测生物制药过程的性能和质量，而不考虑规模、设施或生物反应器系统。这项研究的最终目标是取代传统的分析技术，这些技术既费力又昂贵，并且在商业生物制造操作的现代化和数字化中发挥关键作用。这场数字革命将缩短开发时间，改进过程控制，提高过程产量，最终降低改变生活的治疗药物的成本。

Metabolic profiling of mammalian cultures enables a greater understanding of the physiological state of cells throughout a bioreactor run in addition to monitoring the key amino acid concentrations. This information is invaluable for the development of bespoke amino acid feeds that can be tailored to individual cell lines to enhance productivity and minimise product heterogeneities. However, a major challenge of metabolic profiling is the cost and challenge of generating time-course metabolic profiles for a high number of cell cultures.This project aims to overcome this hurdle by combining the power of the next generation high-throughput (HT) Raman spectroscopy device with an advanced micro-bioreactor system. This HT Raman spectroscopy device requires only a small volume equal to 25-100 micro-L and generates a high-resolution Raman spectra. These spectra enable the prediction of amino acid concentrations in addition to all relevant bioreactor off-line measurements including glucose, lactate, cell density and product concentration which opens up significant opportunities to develop better control strategies. The work will generate a data set to train and apply advanced multivariate data analysis (MVDA) and Machine Learning (ML) models to build correlations between the Raman spectra and all available off-line measurements of interest. These measurements can be incorporated into a control strategy enabling at-line control of essential amino acid concentrations and provides the foundation for the development of customised feeds to enhance bioreactor performance and maintain desired product quality specifications. One of the primary objectives of this research is to develop advanced algorithms to autonomously generate robust mathematical models enabling near real-time predictions of key process parameters to support advanced control decisions and enhance manufacturing operations. These novel computational models will be capable of predicting in silico the performance and quality of biopharmaceutical processes regardless of scale, facility or bioreactor system.This research ultimately aims to replace traditional analysis techniques that can be laborious and costly and play a pivotal role in the modernisation and digitalisation of commercial biomanufacturing operations. This digital revolution will lead to shorter development timelines, improved process control and higher process yields, ultimately reducing the cost of life-changing therapeutic drugs.