Context Quantitative structure–retention relationship is a key method for rapidly predicting the retention time of compounds. However, there have been few systematic studies on the application of QSRR to predict the retention time of phenolic acids, and there are many molecular descriptors and reference standards. This model utilized the experimental chromatographic retention parameter (tR value) of ten phenolic acid compounds as the dependent variable, while the solubility energy of the compounds in water (EW) and in methanol (EM) served as the independent variables. A new QSRR method was established to predict the retention time of phenolic acid compounds, and the QSRR model maintained stable and good prediction performance under a variety of reasonable calculation methods. This study also investigates for the first time the impact of different computational methods under DFT on the accuracy of phenolic acid solvation energy calculations and the predictive performance of QSRR. This research has significantly improved the capability, efficiency, and reliability of theoretical studies, data analysis, and practical applications of phenolic acids throughout our entire field by establishing an optimal computational method. Statistical analysis showed no significant difference in prediction error between models built using the 6-31G basis set and larger, more computationally expensive methods, and the model has been validated to demonstrate good predictive performance. Therefore, using the 6-31G basis set for descriptor calculation is a highly cost-effective choice for phenolic acid QSRR studies. Methods HPLC was used to obtain the retention times of the compounds. GaussView 5.0 and Gaussian 09W were employed to perform structural optimization and molecular descriptor calculations for the corresponding compounds using different methods. Stepwise multiple linear regression fitting was then used to calculate the absolute values of the errors in the chromatographic retention parameters for each model. Paired sample t-tests were subsequently conducted to compare the effects of using different methods to calculate solubility on the model performance. The structure optimization was performed employing the DFT-RB3LYP method from the Gaussian09W package, with various calculation settings including 6-31G, 6-31G-D3, 6-31 + + G-D3, 6-31G, 6-31G-D3, 6-31G, 6-31G-D3, 6-31 + + G**-D3, 6-311G, 6-311G-D3, 6-311 + + G-D3, 6-311G, 6-311G-D3, 6-311G, 6-311G-D3, and 6-311 + + G (a total of 16 different methods) that were calculated. The solubility energies (EW and EM) of the phenolic acids in water/methanol were calculated using the DFT-RB3LYP 6-311 + + G/methods for structural optimization.

Influence of Gaussian calculation method settings on QSRR model accuracy for DFT-calculated phenolic acid solubility energy

期刊论文
发表日期：

2025年11月11日

查看全文

Hongyue Li,
Shiyuan Sun,
Zhou Fang,
Danrong Ni,
Xinran Zhang,
Shuting Zhang,
Shushuang Shen,
Changhai Sun,
Liting Mu

摘要

Context

Quantitative structure–retention relationship is a key method for rapidly predicting the retention time of compounds. However, there have been few systematic studies on the application of QSRR to predict the retention time of phenolic acids, and there are many molecular descriptors and reference standards. This model utilized the experimental chromatographic retention parameter (t_R value) of ten phenolic acid compounds as the dependent variable, while the solubility energy of the compounds in water (E_W) and in methanol (E_M) served as the independent variables. A new QSRR method was established to predict the retention time of phenolic acid compounds, and the QSRR model maintained stable and good prediction performance under a variety of reasonable calculation methods. This study also investigates for the first time the impact of different computational methods under DFT on the accuracy of phenolic acid solvation energy calculations and the predictive performance of QSRR. This research has significantly improved the capability, efficiency, and reliability of theoretical studies, data analysis, and practical applications of phenolic acids throughout our entire field by establishing an optimal computational method. Statistical analysis showed no significant difference in prediction error between models built using the 6-31G basis set and larger, more computationally expensive methods, and the model has been validated to demonstrate good predictive performance. Therefore, using the 6-31G basis set for descriptor calculation is a highly cost-effective choice for phenolic acid QSRR studies.

Methods

HPLC was used to obtain the retention times of the compounds. GaussView 5.0 and Gaussian 09W were employed to perform structural optimization and molecular descriptor calculations for the corresponding compounds using different methods. Stepwise multiple linear regression fitting was then used to calculate the absolute values of the errors in the chromatographic retention parameters for each model. Paired sample t-tests were subsequently conducted to compare the effects of using different methods to calculate solubility on the model performance. The structure optimization was performed employing the DFT-RB3LYP method from the Gaussian09W package, with various calculation settings including 6-31G, 6-31G-D3, 6-31 + + G-D3, 6-31G*, 6-31G*-D3, 6-31G**, 6-31G**-D3, 6-31 + + G**-D3, 6-311G, 6-311G-D3, 6-311 + + G-D3, 6-311G*, 6-311G*-D3, 6-311G**, 6-311G**-D3, and 6-311 + + G** (a total of 16 different methods) that were calculated. The solubility energies (E_W and E_M) of the phenolic acids in water/methanol were calculated using the DFT-RB3LYP 6-311 + + G**/methods for structural optimization.