Joep Vanlier Email & Phone Number

Amsterdam Area, Netherlands

Eindhoven Area, Netherlands

During my PhD I have worked on propagating uncertainty in biochemical computer modelscomprised of ordinary differential equations. After gaining experience with frequentist and Bayesian approaches to uncertainty analysis, I have come up with a strategy to reliably infer predictive distributions (by compiling Profile Likelihood with geometry based Markov Chain Monte Carlo sampling). This strategy enables simulation of predictive distributions that reflect parameter uncertainty. When I first started my PhD, most of the methodological development was focused on determining the parameter uncertainty distribution. The main contribution I have made to the field is to use these predictive distributions for experiment selection by exploiting the fact that both the model structure and data imply specific non-linear relations between samples from these distributions. I have developed two novel approaches that exploit the relations between these predictive distributions. The first method is geared towards reducing the uncertainty of selected predictions of interest (Bioinformatics 2012), while the second method is aimed at designing experiments to allow for successful model selection (currently under review). The major advantage of using these predictive distributions over the commonly used linearizations and asymptotic approximations is that they work well when data is scarce. Furthermore, they are generally applicable since predictive distributions can be computed for a wide range of quantities. If it can be simulated, it can be targeted or used as potential measurement. This is particularly useful when dealing with models of biochemical networks, since directly measuring states is often not possible.

Doctor Of Philosophy (Phd), Biomedical Engineering

Msc, Biomedical Engineering, Cum Laude

Bsc, Biomedical Engineering