The cellular network and its environment govern cell and organism behavior and are fundamental to the comprehension of function, misfunction and drug discovery. Over the last few years, drugs were observed to often bind to more than one target; thus, poly-pharmacology approaches can be advantageous, complementing the "one drug - one target" strategy. Targeting drug discovery from the systems biology standpoint can help in studies of network effects of mono- and poly-pharmacology. In this mini-review, we provide an overview of the usefulness of network description and tools for mono- and poly-pharmacology, and the ways through which protein interactions can help single- and multi-target drug discovery efforts. We further describe how, when combined with experimental data, modeled structural networks which can predict which proteins interact and provide the structures of their interfaces, can model the cellular pathways, and suggest which specific pathways are likely to be affected. Such structural networks may facilitate structure-based drug design; forecast side effects of drugs; and suggest how the effects of drug binding can propagate in multi-molecular complexes and pathways. Cellular functions are performed through protein-protein interactions; therefore, identification of these interactions is crucial for understanding biological processes. Recent studies suggest that knowledge-based approaches are more useful than "blind" docking for modeling at large scales. However, a caveat of knowledge-based approaches is that they treat molecules as rigid structures. The Protein Data Bank (PDB) offers a wealth of conformations. Here, we exploited an ensemble of the conformations in predictions by a knowledge-based method, PRISM. We tested "difficult" cases in a docking-benchmark data set, where the unbound and bound protein forms are structurally different. Considering alternative conformations for each protein, the percentage of successfully predicted interactions increased from about 26 to 66%, and 57% of the interactions were successfully predicted in an "unbiased" scenario, in which data related to the bound forms were not utilized. If the appropriate conformation, or relevant template interface, is unavailable in the PDB, PRISM could not predict the interaction successfully. The pace of the growth of the PDB promises a rapid increase of ensemble conformations emphasizing the merit of such knowledge-based ensemble strategies for higher success rates in protein-protein interaction predictions on an interactome scale. We constructed the structural network of ERK interacting proteins as a case study. We constructed and simulated a "minimal proteome" model using Langevin dynamics. It contains 206 essential protein types that were compiled from the literature. For comparison, we generated six proteomes with randomized concentrations. We found that the net charges and molecular weights of the proteins in the minimal genome are not random. The net charge of a protein decreases linearly with molecular weight, with small proteins being mostly positively charged and large proteins negatively charged. The protein copy numbers in the minimal genome have the tendency to maximize the number of protein-protein interactions in the network. Negatively charged proteins that tend to have larger sizes can provide a large collision cross-section allowing them to interact with other proteins; on the other hand, the smaller positively charged proteins could have higher diffusion speed and are more likely to collide with other proteins. Proteomes with random charge/mass populations form less stable clusters than those with experimental protein copy numbers. Our study suggests that "proper" populations of negatively and positively charged proteins are important for maintaining a protein-protein interaction network in a proteome. It is interesting to note that the minimal genome model based on the charge and mass of Escherichia coli may have alarger protein-protein interaction network than that based on the lower organism Mycoplasma pneumoniae. Proteins function through their interactions, and the availability of protein interaction networks could help in understanding cellular processes. However, the known structural data are limited and the classical network node-and-edge representation, where proteins are nodes and interactions are edges, shows only which proteins interact; not how they interact. Structural networks provide this information. Protein-protein interface structures can also indicate which binding partners can interact simultaneously and which are competitive, and can help forecasting potentially harmful drug side effects. Here, we use a powerful protein-protein interactions prediction tool which is able to carry out accurate predictions on the proteome scale to construct the structural network of the extracellular signal-regulated kinases (ERK) in the mitogen-activated protein kinase (MAPK) signaling pathway. This knowledge-based method, PRISM, is motif-based, and is combined with flexible refinement and energy scoring. PRISM predicts protein interactions based on structural and evolutionary similarity to known protein interfaces. We focus on improvement of PRISM toward modeling of specific interaction of key proteins in cancer and inflammation pathways to figure out regulation in central processes in the cell.