Network Robustness

Percolation analysis reveals different network substructure

Protein and proteoform networks present different substructure in the clusters formed by the nodes. This can be easily observed by the percolation curve plots. We can not conclude that network analysis methods will work with the same efficacy on both types of networks. Different network analysis methos should be devised for analysing samples in these reference networks.

Proteoform networks

Proteoform network nodes have smaller average degree. We perfomed node and link percolation analysis by removing random connections or nodes from the network to observe if the network stays connected. Proteoform networks showed to be less robust requiring to keep a bigger number of connections and nodes to keep showing the giant component (keep majority of nodes connected). This is a natural result from the fact that protein networks are more robust due to more backup connections per node to stay connected to the rest, even when the connections are less specific and biologically significant.

We remember that the biological networks we have at hand contain a subset of the nodes (proteins, proteoforms) and links (connections) of the real biological processes happening in the nature. We still perform analysis for different biological processes or diseases. Disease modules are subnetworks that encompass the nodes and links of the proteins demonstrated to be associated with the disease or biological process of interest. Disease modules in the protein network are easier to observe because the nodes belonging to the disease module might be connected to each other since early stages. For the case of proteoforms, since the connections are fewer for each node and more specific, we require more annotations in the databases to actually observe the disease modules.