A combined bi-objective optimization and Bayesian framework to postulate pharmacometric compartmental models

Abstract

Recently, some applications of Process Systems Engineering to physiology and clinical medicine make use of compartmental analysis to represent transport of material in biological processes. One of the first steps of this analysis is to generate a set of plausible models that describe the system under study. In a previous work, we have proposed an optimization framework to support this task using a superstructure approach which inherently considers the different feasible flows between any pair of compartments. In this work, we extend such a framework to a bi-objective optimization that allows evaluating the trade-off between model fitness and complexity. To discriminate among the different models in the Pareto frontier, we employ a Bayesian metric which is approximated using a Markov Chain Monte Carlo sampling. We present a case study related to an immuno-oncology agent pharmacokinetics to demonstrate the advantages and limitations of the proposed approach.