Artificial Neural Network Modelling of the Retention of Acidic Analytes in Strong Anion-Exchange HPLC: Elucidation of Structure-Retention Relationships

Abstract

Computational models can be used to increase understanding of physical processes within chromatographic systems, leading to more efficient method development and optimisation strategies. In ion-exchange chromatography, various models have been derived to predict retention time; however, there remains a gap in understanding regarding the elucidation of fundamental processes contributing to retention. Here, artificial neural networks have been used to model retention of simple acidic analytes by strong anion-exchange HPLC in an attempt to understand what other factors aside from simple electrostatic interactions between ionised analyte, stationary phase and counter-ion contribute to the differential elution order of such compounds. The weights assigned by each neuron to the inputs in trained networks were used to infer the influence of a number of physicochemical analyte properties to retention under various conditions. These showed that several retention mechanisms were operating simultaneously, and that the contribution of each varied as eluent ionic strength and composition were altered at constant apparent pH. Analyte pKa had most influence on retention under most conditions, but analyte volume, LogP, and steric and electronic effects were also prominent, especially in eluents containing water.

Prediction of Chromatographic Retention Time in High-Resolution Anti-Doping Screening Data Using Artificial Neural Networks

“…These use solute descriptors and linear regression modeling to predict a retention parameter for a particular set of compounds. Both algorithm-based and predictive ANN approaches have incorporated the use of LSERs as inputs to estimate retention (usually as retention factor, k ) across a number of chromatographic modes, − column formats, − for method optimization purposes, − to predict retention behavior across column formats, or to estimate the retention of unknowns. − Each LSER solute descriptor must be calculated, and each coefficient is usually obtained via measured retention data for a set of representative compounds. Thus, the time to produce an accurate LSER model can be significant (as well as especially challenging for ionizable compounds in particular , ).…”

Characterization of stationary phases based on monosubstituted benzene retention indices using correspondence factor analysis and linear solvation energy relationships in RPLC

“…The coefficients of both LSER-k and LSER-I show that the retention expressed by log k or I depends primarily on the solute molar volume v (positive sign), noting an overall attractive interaction with the stationary phase (increase the retention). For both models, the solute hydrogen bond Table 4 Coefficients of both LSER-log k and LSER-I and their statistics for seven C 18 and one C…”

“…Among the models applied currently, the QSRR model has been widely used for studying different chromatographic systems and predict of primary retention data in LC [7][8][9], particulary, in chemometric methods as principal component analysis (PCA) and correspondence factor analysis (CFA) [10][11][12][13] and in the linear solvation parameter model based on the linear solvation energy relationships (LSER). A QSRR model has been also used for characterizing and comparing of stationary phases [14][15][16] and the elucidating of retention mechanisms in LC [17,18]. The retentions of selected solutes are related to specific interactions by the following equation:…”

Interpretative optimization of the isocratic ion chromatographic separation of anions

“…6 Recently, the computer-assisted IC separation has addressed this problem, using factorial design (FD), 7 8 -9 simplex method 1 or neural network approach. 10,11 Artificial neural network (ANN) modelling has ability to predict separation with no prior knowledge of the retention model. However, without the input of preliminary data, ANN will provide little information useful in improving the quality of separation.…”