Data-Derived Modeling Characterizes Plasticity of MAPK Signaling in Melanoma
Figure 5
Workflow for network definition, fitting and simulation. Definition of starting topology.
The initial step was the implementation of a network topology. Grey nodes are measured in our experimental assay. Blue nodes are not modeled or measured and are depicted here to enable understanding of network as a whole. Edges represent directed regulatory interactions reported in literature, and hence can represent activation or inhibition of the target. Condition-dependent gFIS training. A gFIS for each signal measured was trained to the corresponding dataset independently for each condition acquired. This enabled determination of the parameter set specific for each condition. mtFIS implementation. The condition-dependent parameters were used to create a multi-treatment model, including a naive condition switch to enable choice of parameters in the simulation for each condition. Network simulation. Upstream species could not be fitted to further upstream regulators. Hence, an input node consisting of a mapping function specified the measured value of the upstream species at the simulated time point. Thereby, the propagation of time as a signal was enabled. In turn, the fitted models were evaluated at the upstream-simulated value. The discontinuous black arrow represents the possibility of additional signaling intermediates upstream of each model. Network evolution. Blue dashed arrows indicate the workflow steps that can be repeated for a subset of the acquired data points up to a defined time point. See main text for the simulation resulting for models fitted to 96, 72 and 48 hours.