Related Experiment Videos
Benchmarking of Computational Methods for Creation of Retention Models in Quantitative Structure-Retention
Ruth I J Amos1, Eva Tyteca1,2, Mohammad Talebi1
1Australian Centre for Research on Separation Science (ACROSS), School of Physical Sciences-Chemistry, University of Tasmania , Private Bag 75, Hobart 7001, Australia.
Journal of Chemical Information and Modeling
|October 14, 2017
Summary
A quick molecular modeling method for predicting analyte structures in quantitative structure-retention relationship (QSRR) models yields results comparable to expensive computational methods. For flexible molecules, vacuum calculations are as effective as solvent-corrected ones.
Area of Science:
- Computational Chemistry
- Cheminformatics
Background:
- Quantitative structure-retention relationship (QSRR) models predict analyte retention times by correlating chromatographic parameters with molecular descriptors.
- Accurate prediction of molecular structures, particularly 3D geometry, is crucial for reliable QSRR model performance.
- Errors in structure calculation can propagate into QSRR models, affecting prediction accuracy.
Purpose of the Study:
- To compare the effectiveness of different computational methods for optimizing molecular structures used in QSRR models.
- To evaluate the impact of vacuum versus solvent-corrected calculations on structure optimization.
- To assess the performance of Natural Bond Orbital (NBO) analysis against Mulliken charge calculations.
Main Methods:
- Molecular modeling, semiempirical, and density functional theory (DFT) methods (B3LYP, M06) were used for structure optimization.
- Calculations were performed in both vacuum and with solvent corrections (acetonitrile, water).
- Natural Bond Orbital (NBO) analysis and Mulliken charge calculations were compared.
Main Results:
- A rapid and cost-effective molecular modeling approach for structure determination produced results comparable to computationally intensive methods.
- For molecules with limited flexibility, vacuum or gas-phase calculations demonstrated similar effectiveness to solvent-corrected calculations.
- The inherent errors in descriptor creation methods may influence the observed outcomes.
Conclusions:
- The choice of computational method for structure optimization in QSRR studies can be guided by a balance of accuracy, cost, and computational time.
- For certain applications, simpler methods like molecular modeling may suffice, especially when considering descriptor creation limitations.
- Vacuum calculations can be a viable and efficient alternative to solvent-corrected methods for low-flexibility molecules in QSRR model development.