Modeling an in silico platform to predict chromatographic profiles of UV filters using ChromSimulator

Abstract

The use of Analytical Quality by Design (AQbD) concepts is increasing for rational development of analytical methods of liquid chromatography. To facilitate the use of the AQbD approach, we developed an in silico tool capable of predicting the chromatographic profile of organic UV filters based on theoretical retention and separation predictions, in absence of experimentation, with sufficient assertiveness to support analytical development. For this purpose, we used Design of Experiments (DoE) in conjunction with chemical understanding offered by Quantitative Structure – Retention Relationship (QSRR), combined with Monte Carlo method (MCM). Seven analytes were used to establish the prediction model: benzophenone-3, butyl methoxydibenzoilmethane, ethylhexyl triazone, ethylhexyl dimetyl PABA, ethylhexyl methoxycinnamate, bis-ethylhexyloxyphenol methoxyphenyl triazine, and octocrylene. The model generated by multiple regression analysis, using Minitab 19 software, obtained a determination coefficient (R2) of 99.82% and an adjusted determination coefficient (R2 adj) of 99.80%, in addition, residual values had normal distribution, homoscedasticity, and independence. From the results obtained, the predictors considered in the model were the proportion of ethanol in the mobile phase, mobile phase pH, mobile phase flow rate, column temperature, and molecular descriptors (Wlambda3.unity, ATSc5, and geomShape). The model exhibited high performance in internal and external validation, obtaining coefficients of prediction (R2 pred) and determination (R2) of 99.71% and 99.79%, respectively. The in silico platform demonstrated great potential for predicting profile and chromatographic parameters of organic UV filters, allowing, without experimentation, an overview of the retention behavior of compounds under various chromatographic conditions.