Cardoso Victor Gustavo Kelis, Balog Julia, Sabin Guilherme Post, Hantao Leandro Wang
Universidade Estadual de Campinas, Instituto de Química, Rua Monteiro Lobato 270, Campinas, SP 13083-862, Brasil.
Instituto Nacional de Ciência e Tecnologia em Bioanalítica (INCTBio), Rua Monteiro Lobato 270, Campinas, SP 13083-862, Brasil.
ACS Omega. 2025 Apr 29;10(18):18775-18783. doi: 10.1021/acsomega.5c00404. eCollection 2025 May 13.
Brazil plays an important role in coffee quality assessment since it is the top producer and exporter. New technologies must be developed to increase production and ensure product quality. Thus, this study presents an application of laser-assisted rapid evaporative ionization mass spectrometry (LA-REIMS) to fingerprint more than 800 Arabica coffee samples. These samples were divided into six sensory classes by professional tasters according to the Brazilian official classification. Machine learning algorithms were applied for a better understanding of complex fingerprints, and their performances were compared. Partial least-squares discriminant analysis (PLS-DA) was inferior in its predictive capability compared to support vector machines (SVM) and artificial neural networks (ANN), which achieved up to 100% accuracy. The high sensitivity to distinct sensory classes enabled a tentative identification of spectral signals, such as fatty acids, chlorogenic acids, and phospholipids, which are likely being related to these properties in Arabica coffee for the first time.
巴西在咖啡质量评估中发挥着重要作用，因为它是最大的生产国和出口国。必须开发新技术以提高产量并确保产品质量。因此，本研究展示了激光辅助快速蒸发电离质谱（LA-REIMS）在对800多个阿拉比卡咖啡样品进行指纹识别方面的应用。这些样品由专业品鉴师根据巴西官方分类分为六个感官类别。应用机器学习算法以更好地理解复杂的指纹，并对它们的性能进行了比较。与支持向量机（SVM）和人工神经网络（ANN）相比，偏最小二乘判别分析（PLS-DA）的预测能力较差，后两者的准确率高达100%。对不同感官类别的高灵敏度使得能够初步识别光谱信号，如脂肪酸、绿原酸和磷脂，这首次表明它们可能与阿拉比卡咖啡的这些特性有关。
