K-cores analysis (Hagmann et al., 2008; Modha and Singh, 2010) is an intermediate scale graph theoretic metric, calculated using directed binarized graphs (Rubinov and Sporns, 2010a,b). K-cores of a graph are a set of connected components that remain, after all vertices of degree less than k have been removed, in an iterative manner. Coreness of node quantifies the highest k-core network a given node belongs to (Fig. 3A). The coreness values of the nodes were evaluated as follows (Shin et al., 2016): First, the 1-core network was identified by finding the isolated 0-degree nodes of the graph. These nodes were given a coreness value of 0, and then deleted from the network to reveal the 1-core network. Next, the 2-core network was identified, and the nodes deleted at this step were given the coreness value of 1. This process was repeated until every node was given a coreness value, until the largest k-core subnetwork of each graph Gw was found. The coreness of all nodes of a patient versus time windows has been plotted as a heatmap, as shown in Figure 5B. This revealed which set of nodes were involved in highly connected subnetworks, and in which time windows this occurred.