Chemical Characterization and Chemometric Modelling of Artemisia dracunculus L. Essential Oil and Hydrolate, Using QSRR-ANN Model

Artemisia dracunculus (tarragon) is an aromatic plant whose essential oil has been extensively studied, while data on its hydrolate are limited. This study characterized the volatile profiles of A. dracunculus essential oil and hydrolate obtained simultaneously by steam distillation and predicted chromatographic retention indices using a combined QSRR–GA–ANN approach. Volatile compounds were analyzed by GC–MS, and molecular descriptors were calculated using PaDEL-Descriptor. Correlation analysis was used to assess descriptor intercorrelations, and a GA was used to select relevant descriptors, followed by the development of a multi-layer perceptron ANN for RI prediction. A total of 47 volatile compounds were detected, with 26 identified in the essential oil (98.8%) and 28 in the hydrolate (97.8%). Methyl eugenol (51.9%) and sabinene (30.6%) were the major constituents of the essential oil, whereas methyl eugenol (51.1%) and terpinen-4-ol (35.3%) predominated in the hydrolate. Seven molecular descriptors were selected for modelling, and the MLP 7–6–1 network showed high predictive performance, with R2 values of 1.000, 0.983, and 0.998 for the training, test, and validation sets, respectively. MATS5e and MATS3p were the most influential descriptors. This chemometric approach may complement experimental GC–MS analysis by supporting RI prediction and interpretation of volatile profiles in A. dracunculus essential oil and hydrolate.