Institute for Instrumental Analytics and Bioanalytics, Mannheim University of Applied Sciences, Paul-Wittsack-Straße 10, 68163 Mannheim, Germany; Institute of Food Chemistry, Hamburg School of Food Science, University of Hamburg, Grindelallee 117, 20146 Hamburg, Germany; Micro-Biolytics GmbH, Schelztorstraße 54, 73728 Esslingen am Neckar, Germany.
Institute for Instrumental Analytics and Bioanalytics, Mannheim University of Applied Sciences, Paul-Wittsack-Straße 10, 68163 Mannheim, Germany.
Spectrochim Acta A Mol Biomol Spectrosc. 2021 Dec 15;263:120225. doi: 10.1016/j.saa.2021.120225. Epub 2021 Jul 23.
In this study, highly reproducible MIR spectroscopy and highly sensitive MALDI-ToF-MS data were directly compared for the metabolomic profiling of monofloral and multifloral honey samples from three different botanical origins canola, acacia, and honeydew. Subsequently, three different classification models were applied to the data of both techniques, PCA-LDA, PCA- kNN, and soft independent modelling by class analogy (SIMCA) as class modelling technique. All monofloral external test set samples were classified correctly by PCA-LDA and SIMCA with both data sets, while multifloral test set samples could only be identified as outliers by the SIMCA technique, which is a crucial aspect in the authenticity control of honey. The comparison of the two used analytical techniques resulted in better overall classification results for the monofloral external test set samples with the MIR spectroscopic data. Additionally, clearly more multifloral external samples were identified as outliers by MIR spectroscopy (91.3%) as compared to MALDI-ToF-MS (78.3%). The results indicate that the high reproducibility of the used MIR technique leads to a generally better ability of separating monofloral honeys and in particular, identifying multifloral honeys. This demonstrates that benchtop-based techniques may operate on an eye-level with high-end laboratory-based equipment, when paired with an optimal data analysis strategy.
在这项研究中，我们直接比较了高度重现的 MIR 光谱和高度敏感的 MALDI-ToF-MS 数据，以对来自三个不同植物学起源（油菜、刺槐和甘露蜜）的单花蜜和多花蜜样本进行代谢组学分析。随后，我们将三种不同的分类模型应用于两种技术的数据，即 PCA-LDA、PCA-kNN 和软独立建模分类分析（SIMCA），作为分类建模技术。使用两种数据集，PCA-LDA 和 SIMCA 均正确地对所有单花蜜外部测试集样本进行了分类，而 SIMCA 技术只能将多花蜜测试集样本识别为离群值，这是蜂蜜真实性控制的一个关键方面。两种分析技术的比较表明，对于单花蜜外部测试集样本，MIR 光谱数据的总体分类结果更好。此外，与 MALDI-ToF-MS（78.3%）相比，MIR 光谱法（91.3%）明显更能将更多的多花蜜外部样本识别为离群值。结果表明，所使用的 MIR 技术具有高度的重现性，这通常导致其能够更好地区分单花蜜，并特别能够识别多花蜜。这表明，当与最佳数据分析策略结合使用时，基于台式的技术可能与高端实验室设备具有同等水平的性能。
