Network-Based Drug Repositioning
Experimental drug development is expensive, labor-intensive, and time-consuming and mostly limited to a smaller number of targets. Current studies reveal that drug repurposing acts more efficiently than de novo experimental approaches as it involves higher costs and risk. Network analysis is a multifaceted platform for drug purposing, as model interactions are constructed from different biological concepts and their corresponding heterogeneous biological networks.
With the recent advancement of technology, network-based analysis in computational biology has attracted increasing attention. It focuses on organizing the relationships among different bio-molecules in the form of networks to extract newly emerged properties at a network level. Under different conditions, these multilevel meta-networks are allowed to investigate the influences of different biological phenotypes on cellular systems.
In the network pharmacology architecture, a network can be viewed as a connected graph. This type of biological networks, are composed of individual molecular entity (e.g., protein, drug, encoded gene, SNP, etc.) and its biological target, potential molecules from target pathways and biological process as node property and, the relationship (direct or indirect interactions, semantic relations, disease association, signal flow) among these nodes are represented as edge properties. System-scale network knowledge is a must require for modern drug discovery. For example, perturbation of biological networks causes complex diseases as the efficacy and the toxicity of a drug are a consequence of the complex interplay among different cellular components.
The prediction drug-target interaction has been recognized several times with core biological network analysis. Although these approaches are lacked by the incompleteness of knowledge curated from multi-omics molecular interactome; based on the availability of the biological data, network-based drug repositioning methods can be grouped into different categories: proteomic and genomic interaction network (PPIN & GGIN), gene regulatory networks (GRN), drug interaction network (DIN), and metabolic networks (several pathways). Additionally, an integrated disease-similarity network from the available gene expression profile can be added as important components of network-based drug repurposing.
The relationship between disease proteins and drug targets in the organism-level protein-protein interaction network leads to a rational, network-based drug purposing design strategy. In general, disease proteins are not scattered randomly in the interactome (organism level) but constructs smaller disease modules such as localized neighborhoods of those disease proteins. The disease protein within the disease module eventually helps to explore the mechanism-of-action of effective drug combinations. The shortest path lengths based on network proximity are captured between drug targets and disease proteins from those networks.
Molecular perturbations occur due to drug administration or disease that can be extracted as expression data for network-based analysis. Such expression level information is exploited to construct GRN or sometimes used to prioritize nodes on existing biological networks, to identify potential candidate genes for drug repurposing. In the GRN-based approach, each biological node is characterized with different neighborhood metrics, such as network propagation, neighborhood scoring, interconnectivity, random walk.
The resource flow mechanism within a biological insightful network is described as network propagation. The fold-change of the node itself and its neighbor is computed by neighborhood scoring. The ordering of candidate nodes is mainly evaluated based on their connection with other differentially expressed nodes (genes). A source of experimentally validated drug-target interactions is employed to look for potential repurposing candidates using these network-based properties.
The metabolic network provides a different perspective to drug-repurposing where each node represents chemical compound and metabolites and edges represents the reactions that can be catalyzed by one or more enzymes. These networks are mostly directed networks and the enzymes are considered as the targets for possible therapies.
Drug target interaction (DTI) prediction is one of the common approaches of network-based drug repurposing. In DTI, Interaction networks are represented through a bipartite graph where drugs or targets are represented as nodes, and the experimentally validated interactions are considered as edges. The aim of these approaches is to identify novel links. In order to make an accurate prediction of drug-target pair, the whole interaction network is augmented with several additional similarity measures. .
Several pieces of evidence from prior experiments have shown that similar structural proteins tend to target similar proteins. The accumulated information extracted from the DTI suggests that many drugs have a tendency to interact with additional targets than the predesigned targets.
However, the multi-omic functional networks (MoFN) are reversely correlated with drug-disease modules. In MoFN network, nodes (proteins or genes) are connected by weighted edges which reflect the probability of sharing a common biological function within those networks.
The computational and technological advancement has increased the possibility to shift the paradigm of single-target-drug to multiple-target-drug by exploring multiple disease protein network analysis with the help of multi-omic network-based approach within similar disease modules considering minimum toxicity profiles of the drug.