Identification and quality evaluation of Camellia sinensis cv. Longjing 43 fresh leaves from Enshi with different altitudes based on non-targeted metabolomics combined with chemometrics.

The quality of green tea is closely related to its growing environment, especially the altitude. In the present study, the intrinsic relationship between green tea raw materials and altitude was investigated by using machine learning in chemometric analysis. Forty-six Camellia sinensis cv. Longjing 43 (LJ43) samples from three altitude ranges (U500, 500-799, H800) were collected and the basic component analysis and non-target metabolomics were conducted. Results revealed significant altitude-dependent trends that tea polyphenols, total catechins (EGCG, EC, C, GC, EGC), and caffeine decreased with the increasing altitude, while the total free amino acids include theanine, soluble sugars, ECG, and CG increased. Additionally, seventy-four altitude-specific metabolites were identified by using dual thresholds of VIP ≥ 1 and P < 0.05, and the multiple machine learning algorithms were further employed to classify the altitude of raw materials. On the test set, the Random Forest (RF) model achieved the highest accuracy of 97.5 %, outperforming Support Vector Machine (SVM) at 95.24 % and K-Nearest Neighbors (KNN) at 82.5 %. The RF ensemble structure helped reduce overfitting on high-dimensional metabolomic data and quantify feature importance, facilitating identification of key altitude biomarkers such as 1,2,3-tri-O-galloyl-β-d-glucose and Galloylprocyanidin B4. In contrast, linear SVM struggled with nonlinear patterns, while KNN was adversely affected by the irrelevant features. This study demonstrated that metabolites were closely related to the altitude. Nontargeted metabolomics analysis coupled with machine learning functions as a powerful approach to establish chemical fingerprint maps of raw materials and their environment.