Linking molecular structure to chromatographic behavior: a quantitative structure-retention relationship study of Olea europaea L. essential oil components

Abstract

A robust quantitative structure-retention relationship (QSRR) model was developed to accurately predict the linear retention indices (LRI) of 51 essential oil compounds. Molecular descriptors were calculated using alvaDesc software, and model construction was achieved through a multiple linear regression (MLR) approach. A rigorous variable selection process identified relevant descriptors, resulting in a statistically significant model with strong predictive performance (R² = 0.9533, Q²_LOO = 0.9339, Q²_LMO = 0.9293, RMSE _tr = 55.0581, s = 50.0169). External validation further confirmed the model’s reliability, demonstrating excellent predictive capability (R²_ext = 0.9381, Q²_F1 = 0.9361, Q²_F2 = 0.9354, Q²_F3 = 0.9646, CCC_ext = 0.9663, RMSE_ext = 40.3308). The findings highlight the efficiency of the QSRR-MLR model in predicting retention indices, providing valuable insights into molecular properties influencing compound retention. Additionally, the applicability domain assessment ensured reliable predictions within the studied chemical space. This methodology offers a deeper understanding of chromatographic behavior and presents potential for application to other chemical classes for predictive modeling.

Keywords: QSRR, essential oil, volatile chemicals, prediction set, validation.