Utility of de novo network inference in resolving peptide level discordance. Protein sub-networks constructed using peptide topological overlap show correlated clusters. Taken together, some proteins show conflicts between constituent peptides, where given a variable such as time, some peptides are increasing in abundance and other decreasing (A). To examine this, only proteins with uniquely mapping peptides were used. A protein sub-network is constructed by taking associated peptides, and keeping only edges with topological overlaps in a specified upper quantile (e.g. the upper 20% of all topological overlaps for the protein). In all three data sources, as the edge threshold is raised, the number of connected components with discordant peptides dramatically decreases (B, C). This suggests that inference of protein abundance can be guided by network topology.