Unraveling volatile metabolites in pigmented onion (Allium cepa L.) bulbs through HS-SPME/GC-MS-based metabolomics and machine learning

Kaiqi Cheng ¹ Jingzhe Xiao ¹ Jingyuan He ¹ Rongguang Yang ¹ Jinjin Pei ¹ Wengang Jin ^1* A. M. Abd El-Aty ^2*

• ¹ Shaanxi University of Technology, Hanzhong, Shaanxi, China
• ² Cairo University, Giza, Giza, Egypt

Colored onions are preferred by consumers for their unique aroma, abundant phytochemicals and various biological activities. The total phenols, flavonoids, anthocyanins, carotenoids, and antioxidant activities of three colored onion bulbs were evaluated in this study, and their volatile metabolites were further detected via headspace solid-phase microextraction combined with gas chromatography-mass spectrometry (HS-SPME/GC-MS). The results showed that the major phytochemicals and antioxidant activities varied among colored onion bulbs. A total of 243 volatile metabolites were detected, comprising sulfur compounds (51%-64%), organic acids and their derivatives (4%-19%), and others. Multivariate statistical analysis revealed significant variations in the volatile metabolites of three pigmented onions, and 19 characteristic metabolites were identified as biomarkers. In addition, 33 and 38 feature metabolites were also extracted via the SelectKBest and LASSO methods, respectively, and 38 features selected via Lasso can effectively distinguish the three colored onions via PCA, uniform manifold approximation and projection (UMAP), and the k-means algorithm. Four machine learning models were further used for the classification of colored onions, and the random forest model was the optimum model, with a classification accuracy of 1.00. Furthermore, SHAP analysis confirmed that 20 features could be potential markers for representing the overall metabolites. These results imply that HS-SPME/GC-MS combined with random forest could be a novel and powerful method for extracting important features of volatile metabolites in colored onion bulbs.