CFE 2019: Start Registration
View Submission - CMStatistics
B0524
Title: Parameter estimation in mixed effect models based on ordinary differential equations: An optimal control approach Authors:  Quentin Clairon - ISPED - Université de Bordeaux (France) [presenting]
Chloe Pasin - Department of Pathology and Cell Biology (United States)
Melanie Prague - ISPED - Universite de Bordeaux (France)
Irene Balelli - INSERM U1219 (France)
Rodolphe Thiebaut - Bordeaux University INSERM Vaccine Research Institute (France)
Abstract: A parameter estimation method is presented for nonlinear mixed effect-models based on ordinary differential equations (NLME-ODEs). These models aim to describe the dynamic of a whole population while accounting for the observed variability between subjects. Their relevance for the analysis of biological processes implying a large number of subjects but limited individual measurements during time has already led to the development of parameter estimation methods based on stochastic algorithms. However, these methods generally: 1) do not consider potential model misspecifications; 2) need to estimate initial conditions for each patient or make strong assumptions on their values; 3) show dramatic degradation of their accuracy in presence of poorly identifiable parameters. To face these problems, we propose an original method based on discrete optimal control theory. This procedure incorporates a possible gap between the model describing the population dynamic and the specific individual dynamic. In addition, it is based on a profiled cost function on the initial conditions to avoid their estimation. We compare our approach with other ones on a model proposed to study the antibody concentration dynamics after vaccination against Ebola virus.