基于近红外光谱的虾青素发酵过程中生物量在位监测方法的建立及应用

Establishment and application of an in-situ monitoring method for biomass during astaxanthin fermentation based on near-infrared spectroscopy

摘要:

生物量(biomass)是微生物高密度发酵过程中必须监控的关键指标之一，常用细胞光密度(optical density, OD)来测定。相较于传统的取样离线检测，实时在位监测技术具有无损、反馈快速和污染减少等优点，能够有效调控和优化发酵过程。为实现发酵过程中关键指标的实时反馈，本研究构建了一个基于漫反射近红外光谱仪的在位监测平台，用于虾青素的高密度发酵过程中生物量的监测。通过实施光谱异常值剔除、比较不同的光谱预处理方法以及间隔偏最小二乘法(interval partial least squares, i-PLS)光谱波段分析，结果表明在1 417-1 650 nm波长范围内存在虾青素发酵生物量的特征波段。在此波段基础上建立的生物量动态预测模型交互验证决定系数(determination coefficient of cross validation, R_cv²)和交叉验证均方根误差(root mean square error of cross validation, RMSECV)分别为0.973和9.32。经过3批次发酵的外部验证表明采用i-PLS方法建立的生物量模型在细胞光密度(OD₆₀₀) 2.46-180.50范围内进行监测，OD平均绝对误差(mean absolute error, MAE)为6.28，展现出较高的预测准确性和稳定性。这表明该模型在虾青素高密度发酵过程生物量监测中具有应用前景。

Abstract:

Biomass is one of the key indicators that must be monitored during the high-density microbial fermentation, and it is usually represented by optical density (OD). Compared with the conventional monitoring method based on offline sampling, real-time in-situ monitoring technologies provide advantages such as non-destructive analysis, rapid feedback, and reduced contamination, thereby enabling effective process control and optimization of the fermentation process. To achieve real-time feedback of critical parameters during the fermentation process, this study developed an in-situ monitoring platform based on diffuse reflectance near-infrared spectroscopy to monitor biomass during high-density astaxanthin fermentation. After spectral outlier removal, comparison of different spectral preprocessing methods, and interval partial least squares (i-PLS) spectral band analysis, the characteristic spectral bands for the biomass in astaxanthin fermentation were identified within the wavelength range of 1 417-1 650 nm. A dynamic prediction model for biomass that was established based on this spectral range achieved R_cv² and RMSECV of 0.973 and 9.32, respectively. The validation through three fermentation batches demonstrated that the model developed based on the i-PLS method accurately monitored OD values ranging from 2.46 to 180.50, with an average absolute error of 6.28, showcasing high predictive accuracy and stability. These results indicate that the model presents strong application potential for biomass monitoring in the high-density astaxanthin fermentation.