Division of Biotechnology/IFM, Linköping University, 581 83 Linköping, Sweden.
J Biotechnol. 2010 May 3;147(1):37-45. doi: 10.1016/j.jbiotec.2010.02.023. Epub 2010 Mar 7.
Software sensors for monitoring and on-line estimation of critical bioprocess variables have mainly been used with standard bioreactor sensors, such as electrodes and gas analyzers, where algorithms in the software model have generated the desired state variables. In this article we propose that other on-line instruments, such as NIR probes and on-line HPLC, should be used to make more reliable and flexible software sensors. Five software sensor architectures were compared and evaluated: (1) biomass concentration from an on-line NIR probe, (2) biomass concentration from titrant addition, (3) specific growth rate from titrant addition, (4) specific growth rate from the NIR probe, and (5) specific substrate uptake rate and by-product rate from on-line HPLC and NIR probe signals. The software sensors were demonstrated on an Escherichia coli cultivation expressing a recombinant protein, green fluorescent protein (GFP), but the results could be extrapolated to other production organisms and product proteins. We conclude that well-maintained on-line instrumentation (hardware sensors) can increase the potential of software sensors. This would also strongly support the intentions with process analytical technology and quality-by-design concepts.
用于监测和在线估计关键生物过程变量的软件传感器主要与标准生物反应器传感器（如电极和气体分析仪）一起使用，软件模型中的算法生成所需的状态变量。在本文中，我们建议使用其他在线仪器，如近红外探头和在线 HPLC，来制造更可靠和灵活的软件传感器。比较和评估了五种软件传感器架构：（1）来自在线近红外探头的生物量浓度，（2）通过添加滴定剂的生物量浓度，（3）通过添加滴定剂的比生长速率，（4）来自近红外探头的比生长速率，以及（5）来自在线 HPLC 和近红外探头信号的特定基质摄取速率和副产物速率。该软件传感器在表达重组蛋白（绿色荧光蛋白）的大肠杆菌培养物上进行了演示，但结果可以外推到其他生产生物和产物蛋白。我们得出的结论是，维护良好的在线仪器（硬件传感器）可以提高软件传感器的潜力。这也将强烈支持过程分析技术和质量源于设计的概念。
