**6.4 Network analysis**

Biological networks, an outstanding way of modeling biological entities and their interactions, can supply significant insight into the mechanism action of drugs and drug targets and symptoms of diseases. The models can be used to determine informative associations between genes, chemicals, proteins, phenotypes, and any other biological entities by statistical analysis, computational models, and leveraging graph theory concepts. Based on the data sources, network analysis can be classified into metabolic networks, protein–protein interaction networks, drug–drug interaction networks, drug-side effect association networks, disease–disease interaction networks, and gene regulatory networks. Consequently, bionetworks and their analysis can be used to identify potential therapeutic agents and drug repositioning [49–51].

#### **6.5 Molecular docking**

Studying the ligand-protein interactions at the molecular level has a crucial role in pharmaceutical research. Therefore, the scientific community focused on the exploration of the binding phenomenon over the years. Accordingly, some theories, such as lock and key hypothesis, induced-fit theory, and conformational selection were introduced for the interpretation of ligand–protein interactions [52]. Historically, the refinement of a complex structure by optimization of the separation between the partners was the first description of the docking term in 1970. Molecular docking was first being developed in 1980 to predict the best matching binding mode and the molecular interactions of a ligand to a macromolecular partner through the generation of a number of probable orientations of the ligand inside the protein cavity. The method comprises two interrelated steps including orientations sampling and a scoring function, which are responsible for reproducing experimental binding mode and ranking of prepared complexes [52, 53]. Molecular docking can classify into rigid, semi-flexible, and flexible types, according to the degrees of flexibility of the ligand and receptor. In the rigid docking-like to lock-key theory, both ligand and protein are considered rigid entities and hence, there is no internal degree of freedom. Semi-flexible docking is a molecular docking simulation with flexible ligand and rigid receptors. Thus, all degrees of freedom of ligand are explored. Recently, several online and standalone software such as AutoDock, AutoDock Vina, Molegro Virtual Docker, Gold, Surflex-Dock, GLIDE, FlexX, DOCK, FRED, and so on, have been developed for computing different types of molecular docking. Most available software for molecular docking uses flexible ligands and several are trying to model flexible receptor proteins. In recent years, with promising advancements in optimization and the development of new molecular docking algorithms, numerous publications have been planned for comparing the performance of different molecular docking tools. However, it should be stressed that comparison between molecular docking methods is problematic, due to the dependance on docking performance with classes of the subjected targets. The ability of molecular docking methods to reveal the possibility of enzymatic reactions is a compelling reason for various applications related to computational drug design and repurposing, hit identification, lead optimization, binding site prediction, mechanisms of enzymatic reactions, and protein engineering [54–56]. Since the emergence of COVID-19, several molecular docking-based studies [57–62] have been planned to introduce effective anti-COVID-19 drugs by means of drug repositioning. In **Figure 1**, the main steps of a molecular docking-based drug repurposing study are represented.

#### **Figure 1.**

*Schematic representation of the main steps of a molecular docking-based drug repositioning process. Target identification, ligand preparation, and results interpretation are the three main steps.*
