Dominant negative inhibition data should be analyzed using mathematical modeling--re-interpreting data from insulin signaling

As our ability to measure the complexity of intracellular networks has evolved, it has become increasingly clear that we need new methods for data analysis: methods involving mathematical modeling. Nevertheless, it is still uncontroversial to publish and interpret experimental results without a model-based proof that the reasoning is correct. In the present study, we argue that this attitude probably needs to change in the future. We illustrate this need for modeling by considering the common experimental technique of using dominant-negative constructs. More specifically, we consider published time-series and dose-response data which previously have been used to argue that the protein S6 kinase does not phosphorylate insulin receptor substrate-1 at a specific serine residue. Using a presented general approach to interpret such data, we now demonstrate that the given dominant-negative data are not conclusive (i.e. that in the absence of other proofs, S6 kinase still may be the kinase). Using simulations with uncertainty analysis and analytical solutions, we show that an alternative explanation is centered around depletion of substrate, which can be tested experimentally. This analysis thus illustrates both the necessity and the benefits of using mathematical modeling to fully understand the implications of biological data, even for a small system and relatively simple data.