Bayesian Multilevel Models in Quantitative Structure-Retention Relationships Studies

Pawel Wiczling, Department of Biopharmaceutics and Pharmacodynamics, Medical University of Gdansk, Gdansk, Polska (wiczling@gumed.edu.pl)

In chromatography, multilevel model is a regression model of retention time measurements, in which all chromatographic-specific parameters are themselves given a probability model. The multilevel model when combined with prior knowledge and Bayesian inference allows to quantify the uncertainty for all model parameters and predictions. In this work, the Bayesian multilevel model is applied to isocratic reversed phase high-pressure liquid chromatography data. The proposed model consists of i) the same deterministic equation describing the relationship between retention factor and organic modifier content, ii) covariance relationships relating structure of analyte and chromatographic-specific parameters through Quantitative Structure-Retention Relationships (QSRR), and iii) stochastic components of intra-analyte and inter-analyte variability. The model was implemented in the Stan software that provides full Bayesian inference for continuous-variable models through Markov Chain Monte Carlo methods. The fitting of retention data simultaneously for a large group of analytes allows to obtain a single model that generalizes well to other (not-tested) analytes. It also quantifies the uncertainty around predictions that is essential for decision making. e.g. during method development.

This project was supported by the National Science Centre, Poland (nr. 2015/18/E/ST4/00449)

---

Abstract Reference & Short Personal Biography of Presenting Author

Dr. Wiczling received his PhD in Pharmacy from Medical University of Gdańsk in 2005. Since 2003 he works at the Department of Biopharmaceutcs and Pharmacodynamics,  Medical University of Gdańsk, Poland. His specialties are analytical and theoretical aspects of chromatographic techniques, development of chromatographic methods for simultaneous determination of dissociation constant and lipophilicity, modeling and simulation of drug pharmacokinetics and pharmacodynamics, and application of Bayesian inference techniques in chromatography and pharmacometrics.