Interaction networks .

We downloaded the drug-target interaction network , where two drugs share an edge if they share a physical binding partner, from MANTRA [26].

We visualized the drug target interaction network with NAViGaTOR 2.3.2 [34].

The largest connected component in the drug-target interaction network comprised 72 drugs, which is significantly larger (P < < 0.01) than what would be expected by chance; random sets of 83 drugs in the drug-drug network yield largest connected components with a median size of only 42 drugs (Fig.

We used drug-target data from DrugBank [24] and ChemBank [25] (as provided in MANTRA [26]) to construct a drug-drug interaction network on the set of CMap drugs; two drugs are linked by an edge if they share one or more protein targets (Fig.

We randomly rewired the human protein interaction network , while preserving the same degree distribution as observed in the original network.

For the host-pathogen interaction networks (G 1 and G2), sets U1 and U2 correspond to pathogen proteins, sets V1 and V2 correspond to host proteins, and sets E1 and E2 correspond to interactions between host and pathogen proteins.

If node annotation is provided, the HPIA algorithm defines a similarity between pathogen proteins u1 6 U1 and uz 6 U2 as where SGDV_p and SGDV_ H denote the graphlet degree vector similarity [57] derived from the host-pathogen interaction network and the host-PPI network, respectively; SE3 represents the BLAST expected value (E-value) similarity [58], defined as 1— E-value for E-values g 1 and 0 otherwise; and SGO denotes the GO term annotation similarity calculated using the Iaccard similarity measure [59].

We aligned each pair of host-pathogen interaction networks 30 times and reported the average and standard deviations of the alignment scores over all runs, as well as the best score (SS Table).

Using multiple host-pathogen interaction networks to predict the role of pathogen proteins

Such approaches build on the assumption that protein interaction networks can be viewed as maps in which diseases can be identified with localized perturbation within a certain neighborhood.

To disentangle these complex interactions it is necessary to study genotype-phenotype relationships in the context of protein-protein interaction networks .

In order to evaluate the extent to which such topological community detection algorithms can be used to predict disease modules, we chose three representative, methodologically distinct algorithms that have been successfully applied to identify communities of functionally related proteins (functional modules) in protein interaction networks : (i) A link community algorithm [14], which is based on link-similarities and can also capture hierarchical communities, (ii) the Louvain method, which maximizes a global modularity function [21], and (iii) the Markov Cluster Algorithm (MCL), which detects dense regions based on random flow [24].

molecules or proteins) interact among each other, i.e., the structure of the underlying interaction network .

Other models, known as dynamic models, include the structure of the interaction network and also an equation for each component, which describes how the state of this component changes in time due to the influence of other cell components (e.g.

Since systems whose interaction networks and dynamics are known equally well are rare, current control strategies are based on either the network structure [7, 9, 10, 12, 13] or its dynamics (function) [11, 20—22].