Abstract Different USA-origin cannabis samples were analyzed by GC-FID to quantify all possible cannabinoids and terpenoids prior to their clustering. Chromatographic analysis confirmed the presence of seven cannabinoids and sixteen terpenoids with variable levels. Among tested cannabinoids, Δ9-Tetrahydrocannabinol Δ9-THC and cannabinol CBN were available in excess amounts (1.2–8.0 wt%) and (0.22–1.1 wt%), respectively. Fenchol was the most abundant terpenoid with a range of (0.03–1.0 wt%). The measured chemical profile was used to cluster 23 USA states and to group plant samples using different unsupervised multivariate statistical tools. Clustering of plant samples and states was sensitive to the selected cannabinoids/terpenoids. Principal component analysis (PCA) indicated the importance of Δ9-THC, CBN, CBG, CBC, THCV, Δ8-THC, CBL, and fenchol for samples clustering. Δ9-THC was significant to separate California-origin samples while CBN and fenchol were dominant to separate Oregon-origin samples away from the rest of cannabis samples. A special PCA analysis was performed on cannabinoids after excluding Δ9-THC (due to its high variability in the same plant) and CBN (as a degradation byproduct for THC). Results indicated that CBL and Δ8-THC were necessary to separate Nevada and Washington samples, while, CBC was necessary to isolate Oregon and Illinois plant samples. PCA based on terpenoids content confirmed the significance of caryophyllene, guaiol, limonene, linalool, and fenchol for clustering target. Fenchol played a major role for clustering plant samples that originated from Washington and Nevada. k-means method was more flexible than PCA and generated three different classes; samples obtained from Oregon and California in comparison to the rest of other samples were obviously separated alone, which attributed to their unique chemical profile. Finally, both PCA and k-means were useful and quick guides for cannabis clustering based on their chemical profile. Thus, less effort, time, and materials will be consumed in addition to decreasing operational conditions for cannabis clustering. GRAPHICAL ABSTRACT

中文翻译：

---

根据主要大麻素和萜类化合物的化学特征对美国大麻样品进行综合分类

摘要 使用 GC-FID 分析了不同的美国原产大麻样品，以在聚类之前量化所有可能的大麻素和萜类化合物。色谱分析证实存在七种大麻素和十六种不同水平的萜类化合物。在测试的大麻素中，Δ9-四氢大麻酚 Δ9-THC 和大麻酚 CBN 分别过量 (1.2-8.0 wt%) 和 (0.22-1.1 wt%)。苯酚是最丰富的萜类化合物，范围为 (0.03–1.0 wt%)。测量的化学特征用于对美国 23 个州进行聚类，并使用不同的无监督多变量统计工具对植物样本进行分组。植物样本和状态的聚类对选定的大麻素/萜类化合物很敏感。主成分分析 (PCA) 表明 Δ9-THC、CBN、CBG、CBC、THCV、Δ8-THC、CBL、和 fenchol 用于样本聚类。Δ9-THC 对分离来自加利福尼亚的样品很重要，而 CBN 和 fenchol 在分离来自俄勒冈的样品与其他大麻样品方面占主导地位。在排除 Δ9-THC（由于其在同一植物中的高度可变性）和 CBN（作为 THC 的降解副产物）之后，对大麻素进行了特殊的 PCA 分析。结果表明，CBL 和 Δ8-THC 是分离内华达州和华盛顿州样品所必需的，而 CBC 是分离俄勒冈州和伊利诺伊州植物样品所必需的。基于萜类化合物含量的 PCA 证实了石竹烯、愈创木酚、柠檬烯、芳樟醇和苯酚对聚类目标的重要性。Fenchol 在聚类源自华盛顿和内华达州的植物样本方面发挥了重要作用。k-means 方法比 PCA 更灵活，生成三个不同的类；与其他样品相比，从俄勒冈州和加利福尼亚州获得的样品显然是单独分开的，这归因于它们独特的化学特征。最后，PCA 和 k 均值都是基于其化学特征的大麻聚类有用且快速的指南。因此，除了降低大麻聚类的操作条件外，还将消耗更少的精力、时间和材料。图形概要 除了降低大麻集群的操作条件外，还将消耗材料。图形概要 除了降低大麻集群的操作条件外，还将消耗材料。图形概要