Fig. 5.
A flowchart of the model selection process to create a sparse differential network is shown. a Raw data is sampled and graphical representations of the network are given for disease and control. The samples are different and therefore have varied network representations. b Differential networks are generated using GGM using various λ parameters, λ : {1, …, λM }, and a model selection is done through stability analysis to see which λ gives the most stable network variance. A threshold for stability, β, allows us to choose the optimal λ, shown in green. c The optimal λ, in this case λm is used on the original raw data, to generate the final sparse differential network