基于拉曼技术的单细胞生长检测方法

目的 单细胞生长检测可以更加科学地揭示微生物代谢变化的规律，为后期微生物工程应用提供指导。针对微生物生长应用于食品安全期和最佳食用期的精准检测问题，本文提出一种基于拉曼技术的单细胞生长检测方法。方法 首先，通过同步培养实验采集了枯草芽孢杆菌两个批次共900个单细胞拉曼光谱（SCRS）数据，其中600个用于训练和测试，另一批次300个用于模型验证。其次，基于主成分分析的特征关系矩阵，提出CP-SP特征评估方法以筛选SCRS特征用于模型检测。再基于XGBoost构建检测模型，并应用网格搜索和交叉验证对检测模型进行调优。最后，应用混淆矩阵、ROC曲线评估模型对细胞滞后期、对数期和稳定期的检测准确率、敏感性和特异性。结果 选用CP-SP筛选的第一、第二和第四主成分较特征贡献率前3个主成分的分类性能提高了3.1%，调优后的细胞生长检测模型测试准确率为96.0%，验证准确率为92.3%。结论 基于拉曼技术的单细胞生长检测方法能准确识别单细胞生长状态且具有较高的泛化能力，可为食品安全和保鲜制定精准调控机制提供科学指导。

Single-cell Growth Detection Based on Raman Technology

Objective Single-cell growth detection can more scientifically reveal the rules of microbial metabolic changes and guide later microbial engineering applications. To study the accurate detection of microbial growth during the food safety period and optimal edible period, a single-cell growth detection method based on Raman technology is proposed in this paper.Methods First, a total of 900 single-cell Raman spectroscopy (SCRS) data were collected from two batches of Bacillus subtilis through a simultaneous culture experiment, of which 600 were used for training and testing and the other 300 for model validation. Secondly, based on the feature relationship matrix of principal component analysis, CP-SP feature evaluation method was proposed to screen SCRS features for model detection. Then, a detection model based on XGBoost was built, and grid search and cross-validation were applied to optimize the detection model. Finally, confusion matrix and ROC curve were used to evaluate the detection accuracy, sensitivity and specificity of the model for cell lag phase, log phase and stationary phase.Results The experiment found that the classification performance of the first, second, and fourth principal components screened by CP-SP was improved by 3.1% compared with the first three principal components of the feature contribution rate. The test accuracy of the optimized cell growth detection model was 96.0%, and the verification accuracy was 92.3%.Conclusion The results show that the single-cell growth detection method based on Raman technology can accurately identify the single-cell growth state and has a high generalization ability, which can provide scientific guidance for the formulation of precise regulatory mechanisms for food safety and preservation.