Hamburg School of Food Science, Institute of Food Chemistry, University of Hamburg, Grindelallee 117, 20146 Hamburg, Germany.
Department of Pediatric Hematology and Oncology, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany; Charité - Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Institute of Pathology, Charitéplatz 1, 10117 Berlin, Germany.
Food Chem. 2019 Jul 15;286:475-482. doi: 10.1016/j.foodchem.2019.01.105. Epub 2019 Jan 23.
Prediction of the geographic origin of white asparagus was realized using inductively coupled plasma mass spectrometry (ICP-MS) and machine learning techniques. The elemental profile of 319 asparagus samples originating from Germany, Poland, the Netherlands, Greece, Spain, China and Peru was determined. Using a support vector machine (SVM) combined with nested cross-validation, a prediction accuracy of 91.2% was achieved when classifying the country of origin. Accuracy can be increased up to 98% on subsets of samples with high SVM prediction scores. Most relevant elements for provenance discrimination were lithium, cobalt, rubidium, strontium, uranium and the rare earth elements. In addition, the multi-elemental method provided specific fingerprints of asparagus cultivation sites as German samples could be assigned correctly with an accuracy of 82.6%. Asparagus variety and harvest year had no significant influence on provenance distinction, which further underlines the robustness of this study.
采用电感耦合等离子体质谱（ICP-MS）和机器学习技术对白色芦笋的地理起源进行了预测。对来自德国、波兰、荷兰、希腊、西班牙、中国和秘鲁的 319 个芦笋样本的元素图谱进行了测定。使用支持向量机（SVM）结合嵌套交叉验证，当对原产国进行分类时，可达到 91.2%的预测准确率。在 SVM 预测得分较高的样本子集中，准确率可提高至 98%。对于产地判别最相关的元素有锂、钴、铷、锶、铀和稀土元素。此外，多元分析方法提供了芦笋种植地的具体指纹图谱，德国样本的正确分类准确率达到 82.6%。芦笋品种和收获年份对产地的区分没有显著影响，这进一步强调了本研究的稳健性。
