Quantitative structure retention relationship (QSRR) modelling for Analytes' retention prediction in LC-HRMS by applying different Machine Learning algorithms and evaluating their performance.

In metabolomics, retention prediction methods have been developed based on the structural and physicochemical characteristics of analytes. Such methods employ regression models, harnessing machine learning algorithms mapping experimentally derived retention time (tR) analytes with various structural and physicochemical descriptors, known as Quantitative Structure Retention Relationships (QSRR) models. In the present study, QSRR models have been developed by applying four Machine Learning regression algorithms, i.e. Bayesian Ridge Regression (BRidgeR), Extreme Gradient Boosting Regression (XGBR) and Support Vector Regression (SVR) using both linear and non-linear kernels, all tested and compared for their retention prediction ability on experimentally derived and on publicly available chromatographic data, using Molecular Descriptors to describe the physical, chemical or structural properties of molecules. Various configurations of the available datasets, in terms of the highly-correlated features levels (defined as the maximum absolute value of the Pearson's correlation coefficient calculated between any pair of features) they contained, were analyzed in parallel. This is the first study, to the best of our knowledge, of the effect of collinearity on the performance of QSRR predictive models. In the vast majority of cases studied there was no statistically significant difference in the performance of the generated QSRR predictive models among the specified dataset configurations, indicative of the ability of the selected regression algorithms to effectively handle collinearity. In terms of the individual performance of the selected regression algorithms, no pattern was found where one algorithm (or class of algorithms) stood out significantly relative to the others among the study datasets.

Theory and application of the two-mode gradient elution in liquid chromatography involving simultaneous changes in temperature and mobile-phase composition.

The theory of the dual-mode gradient elution in liquid chromatography involving any type of simultaneous changes in column temperature and mobile-phase composition is developed following Drake's approach. The theory was tested in the retention prediction of six alkylbenzenes in aqueous eluting systems modified by acetonitrile. Significant delay phenomena, i.e., a lag between the programmed gradient temperature and the temperature in the oven, and a lag between the oven temperature and the effective temperature the analyte feels in the column, were detected. When these phenomena were taken into account, the retention prediction obtained for all solutes under all dual-mode gradient conditions was excellent. The average percentage error between experimental and predicted retention times is below to 2%.

Modeling the combined effect of temperature and organic modifier content on reversed-phase chromatographic retention. Effectiveness of derived models in isocratic and isothermal mode retention prediction.

Models considering simultaneously mobile phase organic content and column temperature were developed in this study by an extension of different equations describing the influence of temperature on solute retention. This extension was achieved by two methods: a semi-thermodynamic and a direct combination of equations expressed separately the dependence of the retention upon each of these factors. The above approaches gave a great number of expressions for the logarithm of the solute retention factor in terms of both temperature and organic content in the mobile phase, ln k(T,phi), determined from the dependence of the standard enthalpy of the retention process on T. From the final expressions of ln k(T,phi) we tested only those with the minimum number of adjustable parameters, i.e. those that correspond to a constant standard enthalpy of the retention process. For this test we examined the retention behaviour of a sample of alkylbenzenes in aqueous acetonitrile eluents. These compounds exhibit ln k versus 1/T plots with a very small curvature. We found that a new equation for ln k(T,phi) based on the adsorption model for retention performs better than all the others. The average percentage prediction error ranges from 0.7 to 1.4%.