Sensitivity Analysis of Non-Linear Dynamic Models: Prioritizing Experimental Research

Abstract This study demonstrates the use of (global) sensitivity analysis as a supporting tool for guiding and prioritising experimental research activities. The basic idea is to focus the experimental efforts on those factors identified as the most significant by the sensitivity analysis. Such an approach is ideal for identifying critical regions in the experimental design space (typically multi-dimensional). These critical regions can later be investigated experimentally by means of the well-established Design of Experiments methodology. The sensitivity-analysis based approach promises better use of resources by preventing wastage of valuable experimental time on investigating factors which are not influential on the system behavior. The approach requires a first-principles model of the system on which the sensitivity analysis is performed, and was tested on a dynamic fermentation model describing antibiotic production by Streptomyces coelicolor.