A bayesian approach to hypothesis based experimental design

INTRODUCTION One of the key aims of mathematically modeling biochemical pathways is the understanding the underlying mechanisms that govern a specific system of interest. A computational model can be used to predict (un)measured behavior orsystem responses and formalize hypotheses in a testable manner. An important factor which complicates such analyses is the fact that data is relatively scarce and consequently the modeler is faced with a situation where multiple parameter sets can adequately describe the measurement data to an acceptable degree. When predictions required to test the hypothesis are insufficiently constrained more data will be required. However, it is often not immediately evident which measurement(s) at which specific time point(s) would be most informative. Although several design criteria exist, many of which aimed at effectively constraining specific parameters we focus on designing experiments specifically targeting the variance of quantities of interest that depend (optionally nonlinearly) on model predictions: METHODS In this work we use a Bayesian approach to infer a posterior distribution based on the model and the data, which we subsequently use for simulation. Then self-normalized importance sampling of the Posterior Predictive Distribution (PPD) is performed in order to perform Optimal Experiment Design (OED). RESULTS We proposed a flexible Bayesian approach to perform hypothesis driven experimental design that enables the modeler to design experiments that target specific predictions of interest whilst considering finite measurement accuracy and model uncertainty (crucial considering the fact that biological measurements often exhibit large degrees of uncertainty). Experiment(s) can be optimized for any quantity that can be expressed in terms of the model simulations and/or model parameters. We presented our method by illustrating its use on a model of the JAK-STAT signaling pathway. For this specific model, we show that depending on the quantity of interest, measuring the peak value of a time course can either be very beneficial or hardly provide any additional constraint at all.