Figure 8

Insulin signaling case. Panel (A) shows the model structures of two models of early insulin receptor signaling,${\mathcal{M}}_{i,c}$ and the chosen help model${\mathcal{M}}_{i,b}$. This example was analyzed previously[45] and is of interest, since the rejection of${\mathcal{M}}_{i,c}$ would suggest that the recycling is a necessary mechanism to explain the data. Depicted in panel (B) is the experimental data Z (red error bars), and fits of${\mathcal{M}}_{i,b}$ (blue, dashed line) and${\mathcal{M}}_{i,c}$ (green, solid line). The measured data represent the increased auto-phosphorylation of the insulin receptor in response to 100 nM insulin in isolated primary human adipocytes, as measured by SDS-PAGE and immunoblotting. Panel (C) shows the bootstrapping cloud in the χ ² vs DW plane, when the bootstraps have been generated by${\mathcal{M}}_{i,c}$ (green circles, cloud size = 10⁴). As can be seen, the cloud lies along the axes, and there is no benefit of using a 2D analysis. Panel (D) shows the χ ² vs χ ² scatter plot of the${\mathcal{M}}_{i,c}$ cloud (green circles, cloud size = 10⁴) generated after fitting both models to bootstrap sets from${\mathcal{M}}_{i,c}$. The corresponding χ ²-values for the experimental data (Z) is also plotted (red box). As can be seen, the cloud lies away from the axis, and the experimental data point explores the obtained direction. Panel (E) summarizes the results. As the clouds have indicated, the χ ² vs DW combination does not improve upon the individual tests, but still lies on the border of rejection. The χ ² vs χ ² tests on the other hand perform better than the individual tests, and the LHR is best of all.