Konstantinidis Spyridon, Welsh John P, Titchener-Hooker Nigel J, Roush David J, Velayudhan Ajoy
Dept. of Biochemical Engineering, The Advanced Centre for Biochemical Engineering, University College London, London, U.K.
Biologics Process Research and Development, Merck & Co., Inc., Kenilworth, NJ, USA.
Biotechnol Prog. 2018 Nov;34(6):1393-1406. doi: 10.1002/btpr.2673. Epub 2018 Oct 9.
Recently, a grid compatible Simplex variant has been demonstrated to identify optima consistently and rapidly in challenging high throughput (HT) applications in early bioprocess development. Here, this method is extended by deploying it to multi-objective optimization problems. Three HT chromatography case studies are presented, each posing challenging early development situations and including three responses which were amalgamated by the adoption of the desirability approach. The suitability of a design of experiments (DoE) methodology per case study, using regression analysis in addition to the desirability approach, was evaluated for a large number of weights and in the presence of stringent and lenient performance requirements. Despite the adoption of high-order models, this approach had low success in identification of the optimal conditions. For the deployment of the Simplex approach, the deterministic specification of the weights of the merged responses was avoided by including them as inputs in the formulated multi-objective optimization problem, facilitating this way the decision making process. This, and the ability of the Simplex method to locate optima, rendered the presented approach highly successful in delivering rapidly operating conditions, which belonged to the Pareto set and offered a superior and balanced performance across all outputs compared to alternatives. Moreover, its performance was relatively independent of the starting conditions and required sub-minute computations despite its higher order mathematical functionality compared to DoE techniques. These evidences support the suitability of the grid compatible Simplex method for early bioprocess development studies involving complex data trends over multiple responses. © 2018 The Authors Biotechnology Progress published by Wiley Periodicals, Inc. on behalf of American Institute of Chemical Engineers Biotechnol. Prog., 34:1393-1406, 2018.
最近，一种与网格兼容的单纯形变体已被证明能够在早期生物工艺开发中具有挑战性的高通量（HT）应用中持续且快速地识别最优解。在此，该方法通过将其应用于多目标优化问题得到了扩展。本文展示了三个HT色谱案例研究，每个案例都呈现出具有挑战性的早期开发情况，并包括三个响应，这些响应通过采用期望函数法进行合并。针对每个案例研究，评估了实验设计（DoE）方法的适用性，除了期望函数法之外还使用了回归分析，涉及大量权重以及在存在严格和宽松性能要求的情况下。尽管采用了高阶模型，但这种方法在识别最优条件方面成功率较低。对于单纯形方法的应用，通过将合并响应的权重作为公式化多目标优化问题的输入来避免其权重的确定性指定，从而促进了决策过程。这一点，以及单纯形方法定位最优解的能力，使得所提出的方法在快速提供操作条件方面非常成功，这些操作条件属于帕累托集，并且与其他方法相比在所有输出方面都具有卓越且平衡的性能。此外，尽管与DoE技术相比其具有更高阶的数学功能，但其性能相对独立于起始条件，并且所需的计算时间不到一分钟。这些证据支持了与网格兼容的单纯形方法适用于涉及多个响应复杂数据趋势的早期生物工艺开发研究。© 2018作者 生物技术进展 由Wiley Periodicals, Inc.代表美国化学工程师学会出版 生物技术进展，34:1393 - 1406，2018。
