Figure 3

Protein co-expression networks are shown to be approximately scale-free. As the soft thresholding power, β, grows, the resulting adjacency matrix increasingly fits the scale-free model. This trend is robust to missing data, shown here with networks constructed using signed similarity matrices. Subsets of peptide data were taken to eliminate missing data (dark lines). Then, incrementally, additional peptides were included containing between one to ten missing data points, shown here with lighter, broken lines. For network analysis, it is strongly in our interest to incorporate peptides with missing data, since it increases the proteome coverage without weakening the model.