Microbial individual-based models and sensitivity analyses: local and global methods

A microbial Individual-based Model (IbM) to deal with yeast populations growing in liquid batch cultures has been designed and implemented in a simulator called INDISIM-YEAST. Interesting qualitative results have already been achieved with its use in the study of fermentation profiles, small inocula dynamics and lag phase, among others. Nevertheless, in order to improve its predictive capabilities and further development, a deeper comprehension of how the variation of the output of the model can be apportioned to different sources of variation must be investigated. One way to consider a sensitivity analysis for this IbM, providing an understanding of how the model response variables react to changes in the inputs, is the statistical study of well-designed computer experiments. The aim of this contribution is to show how the insights into nine individual cell parameters of INDISIMYEAST, mainly related to uptake and reproduction sub-models, can be obtained by combining local and global sensitivity analyses using simple and classic methods. From data obtained with an extensive set of computer experiments, a study of the variability observed in the evolution of two outputs of this model, ethanol production and mean biomass of the population, was performed. In addition, mono-factorial (one-at-a-time) analyses and ANOVA-based global analyses were also carried out on these two outputs. The model is clearly less sensitive to some parameters than others, depending on the output controlled. Moreover, this study allows identification of the parameters which have the greatest impact on the corresponding outputs and their significant first-order interactions. This work must be understood as an exercise to set up the procedure to be used in a sensitivity analysis study involving microbial IbMs. The knowledge gained will facilitate future parameterization and calibration of different parameters and outputs depending on the purpose of any study