The availability of therapeutically important protein 3D structures made SBDD the most desirable approach for drug design and development ahead of ligand-based drug design (LBDD). However, to enhance the success rate of recent drug development initiatives, it has become customary to integrate SBDD with LBDD approaches in a complementary manner. Using the 3D structures of targets to study and exploit the catalytic pocket, SBDD can search and create appropriate ligands that can subsequently be verified and optimized experimentally. To mitigate the difficulties associated with experimental techniques, several types of SB and LB tactics, or a mix of them, might be used at various phases of TB drug design and development. With the availability of TB multi-omics and a large amount of structural biodata, we can use cheminformatics data mining, data engineering, docking, and homology modeling to identify potential targets.

Virtual screening facilitates choosing the most promising prospective ligand(s) from a database comprising millions of identified molecules for a specific tuberculosis target. The output of candidate compound validation using structure-activity (SA) studies enables a better understanding of the mechanism of action and ADMET (absorption, distribution, metabolism, excretion, and toxicity) properties, thus allowing better development of compounds with improved activity and better pharmacological profiles. The information gathered from this research (both good and negative outcomes) may be saved and used for additional iteration and technique optimization in the future design of new tuberculosis drugs. More than 800 CADD software and webservers (free or commercial) are available, hosted by the Swiss Institute of Bioinformatics at www.click2drug.org (accessed on 5 September 2021). These provide unlimited opportunities to explore drug discovery and design [1]. Table 5 summarizes some available CADD software.

SBDD takes advantage of target protein 3D structure availability. However, if the 3D model of the therapeutically important receptor is not available, computational approaches through homology modeling enable the 3D model prediction of the receptor. Homology modeling or comparative modeling is the most reliable method for 3D protein structure prediction. The methodology entails predicting the 3D structure of the receptor from a homologous protein with at least a 40% similarity index. Threading and ab initio modeling are also methods of protein structure prediction [15]. After obtaining the 3D structure of the target, it is crucial to validate the model by examining the molecular characteristics in a Ramachandran plot. This metric shows the distribution of the ϕ and ψ dihedral angle conformations of the constituting residues in the receptor structure [179]. There are several techniques to validate the predicted protein model [15,175,180].

After determining the target structure, the next step is to determine the catalytic pocket. Catalytic or binding pockets are tiny spaces where ligands attach to the target, inducing the intended result. Consequently, it is crucial to identify the most suitable location on the target protein for ligand binding. Even though protein is dynamic in nature, only a few techniques can identify possible binding residues in the binding pocket. Identification of binding sites on a particular target requires the knowledge of interaction energy and van der Waals (vdW) forces. There are many strategies for catalytic site mappings using interaction energy computation through SBDD. This technique can identify locations on the target receptor that interact positively with functional moieties on the drug-like compounds. These approaches find probes that have energetically advantageous interactions with proteins. Q-SiteFinder [176] is an energy-based technique for predicting catalytic sites widely utilized in the pharmaceutical industry. It is possible to compute the vdW interaction energies of proteins with a methyl probe by using this approach. Those with favored energy values are maintained and grouped in the final product. The total interaction energies of these probe clusters serve as the determinant of their ranking. Aside from that, the functional annotation of interacting protein residues in the binding site allows for the determination of the binding site.

It is also essential to remember that additional possible binding sites, referred to as allosteric sites, may also be present on the target protein surface. Drug development attempts in the conventional sense frequently target the important (orthosteric) binding site to prevent natural substrate binding. Besides, researchers have unveiled noncatalytic sites of Mtb proteins. Shi and colleagues identified a second druggable binding site (allosteric) in Mtb UDP-galactopyranose mutase (UGM) [181]. MS-208, a well-known Mtb-UGM inhibitor, was categorized as a noncompetitive/mixed inhibitor based on NMR and kinetics investigations. This observation implies that MS-208 binds to another location in the receptor and affects the natural enzyme substrate from recognizing the primary pocket. They [181] predicted the allosteric sites for MS-208 on the enzyme via docking with AutoDock Vina [182]. The two identified regions, designated A-site and S-site, show favorable and stable interaction with the ligand after molecular dynamics using Amber [183]. Simulations facilitate structural and functional relationship determination. Because the A-site-bound structure demonstrated the most stable complex formation with good interaction energy and a higher number of contacts, they hypothesized that this site represents an allosteric druggable binding site in Mtb-UGM [181].

After appropriate identification of all druggable sites on the receptor, next comes hit discovery, accomplished by docking chemical libraries into the active cavity of the target receptor. Earlier, the routine in lead discovery required choosing a specific collection of ligands that can play a critical role in identifying and optimizing leads [177]. SBDD blends two distinct approaches for hit search (VS and de novo design) into a single framework.

Currently, vs. has emerged as a dynamic and profitable technique in the pharmaceutical business, particularly for prospecting new drug-like compounds or so-called lead identification [184]. There are two forms of VS: ligand-based vs. (LBVS) and structure-based vs. (SBVS). Biological data is processed in LBVS to distinguish inactive molecules from active ones. Based on consensus pharmacophores, this information facilitates highly functional scaffold identification [185], similarity, or various descriptors. LBVS produces results that are closely related to known active pharmaceutical ingredients. The procedure involves scanning chemical libraries of structures to find molecules with known like potency or that share a pharmacophore or moiety with known activity. The results are typically positive (pharmacophore substructure similarity search) [186]. A moiety substructure search requires using the 2D- or 3D-structure of various ligands to find closely related structures. Usually, comparable substances have similar effects when using ligand-based techniques; thus, they are called similarity methods. For example, if one or more active compounds are known, it is feasible to search a database for comparable but more potent compounds [187].

SBVS allows docking numerous chemical compounds against an enzyme-binding or catalytic site in a short time [188,189,190,191]. The computer algorithms facilitate target protein docking with one of the vast libraries of drug-like chemicals that are commercially or publicly accessible (Table 4). Subsequent steps for search refinement are molecular docking, MD simulations, and experimental tests to obtain IC₅₀ or other efficacy parameters [192]. SBVS relies on the scoring of ligands to function correctly. In contrast to ligand-based techniques, structure-based techniques do not rely on previously collected experimental data to be effective.

De novo drug design involves creating unique chemical compounds from the ground up, starting with molecular building blocks. The essence of this technique is to design chemical structures of tiny molecules that bind to the target active site with high affinity [193], then test these structures experimentally. A variation in approach is typically employed when designing from scratch, and the design algorithm must integrate the search space information acquired. Usually, researchers incorporate positive and negative designs with one another. When using the former strategy, a search is to constrain certain regions of chemical space, which increases the likelihood of discovering results with notable characteristics. The search parameters are set in the negative mode to avoid choosing false positives [194]. Despite its sufficiency in functional scoring analysis, chemical compound design using computational approaches connects organic synthesis but cannot replace it [195]. It is fundamental in the design stage to conduct a thorough evaluation of candidates’ compounds. One of these evaluation tools is the scoring function; multiple scoring functions for multi-objective drug discovery hybrids [196] create many different characteristics simultaneously.

De novo drug design is in two categories: (A) ligand-based drug design and (B) receptor/enzyme-based drug design. The latter method is popular currently. Creating appropriate small molecules for enzyme-based design requires high-quality target protein structures and precise knowledge of proteins’ active sites. The approach entails small molecules designed by matching fragment moiety into the target proteins’ binding pockets. The process requires using computer programs or co-crystallization of the ligand with the receptor [197]. Two ways to execute the ligand-based design are by linking together crucial components, such as atoms or fragments (single rings, amines, and hydrocarbons) to produce an entirely new chemical molecule, or by simply generating ligands from a single parent unit. The fragment-linking technique uses information of the active site to map the likely interaction locations for the different functional groups contained in the design drug fragments [198,199]. One must link these functional groups’ moieties to one another to form an absolute compound. The fragment-growing method features fragment development within the active site, monitored by appropriate search algorithms [199].

These search algorithms make use of scoring systems to determine the likelihood of growth. Fragment-based de novo design is a method of creating new molecules that use the whole chemical space. When using the linking technique, the selection of linkers is imperative. The outside-in strategy and the inside-out approach are both methods for anchoring fragments in the binding site. The outside-in system is the more common method. The outside-in methodology involves the construction blocks placed near or on the edge of the binding site, and the active site gradually expands inward. The inside-out method uses construction pieces randomly placed within the active site region, then constructed outward [100].

Molecular docking has been a prime computational technique of SBVS against Mtb enzymes. The molecular-docking technique was the subject of many published research articles as a tool in drug design against Mtb (Table 6). According to the Himar1 transposon mutagenesis study conducted by DeJesus in 2017 [200], the majority of the enzymes targeted by this method are enzymes encoded by crucial genes, with the exclusion of antigens BioA, NarL, 85c, EthR, and LipU. Although this technique assigns nonessentiality to genes based on in vitro growth, it cannot be relied on to determine whether genes are nonessential in vivo [201,202]. For instance, the NarL enzyme is necessary for anaerobic survival throughout infection, while BioA is crucial for biotin synthesis during the latency phase of Mycobacterium TB infection [203,204]. Furthermore, the EthR protein functions in developing ethionamide resistance and, consequently, survives potentials after drug treatment [205,206].

Many studies of the different targeted enzymes indicate that many are engaged in either intermediate metabolism or lipid metabolism in Mtb. In addition, DNA and RNA regulatory enzymes and cell wall regulator proteins make up the remaining target proteins. Researchers show at least three SBVS efforts against DprE1 and InhA, with InhA being the most frequently targeted. Studies show that InhA is the ultimate target of both isoniazid and ethionamide once activated [218]. As a result, InhA provides a validated target whose suppression has an in vivo influence on the survival of Mtb. Also, numerous antimycobacterial medicines target DprE1 in the current anti-TB research pipeline [29,59]. PyrG, a newly confirmed TB target, also attracted the attention of researchers [219]. The chemicals used in most research works (Table 6) are from generic chemical databases containing millions of compounds, while TB-specific databases and natural product, therapeutic repurposing-focused, and other libraries comprise other chemical compounds reported. As a result, the focus of these early drug discovery initiatives continues to be on totally new drug-like chemical discoveries. The apparent lack of further experimental evidence (in vitro or in vivo) showing compound bioactivity in many of these investigations (Table 6) is an evident issue that precludes these anticipated compounds from being carried onward [206].

Another important computational tool in drug discovery is the density functional theory (DFT), which applies to TB research for the investigations of catalytic processes [220,221], structure-activity relationship analysis [222], and inhibitor potency [192,223]. Chi and colleagues [223] adopted DFT in an anti-tubercular study to confirm their first observations of a change in an inhibitor-binding mechanism in the MbtI protein after adding a substituted enolpyruvyl moiety to the parent chemical structure previously generated from isochorismate. From their [223] observation, there were two distinct binding mechanisms (states 1 and 2) noted in the X-ray crystal structures of MbtI complexed with its inhibitors, indicating that the active site is flexible enough to permit ligand binding. With the aid of Gaussian 09 software application [224] and a theoretical-level hybrid B3LYP [225,226], they computed the global minimum configuration of the (E)-3-(1-carboxyprop-1-enyloxy)-2-hydroxybenzoic acid (AMT), Z-methyl-AMT, and E-methyl-AMT inhibitors complexed in solution. The results revealed that the global minimum geometry of both free Z- and E-methyl-AMT is comparable to its bound geometry (state 2), showing that its arrangement enables binding to MbtI. The computation of conformational entropy quantities for the three molecules indicated that Z-methyl-AMT is the least disorganized. Z-methyl-AMT has a conformational lock provided by the methyl moiety in its structure. Even though a pure Z-isomer has not yet surfaced to distinguish it from the E-isomer empirically, this discovery [224] justifies the powerful interaction of methyl-AMT to MbtI. It provides further knowledge for the future creation of new and effective MbtI drug-like compounds with the aid of DFT.

Despite the widespread success and popularity of DFT, it has flaws stemming from the approximations employed in its operational mode. DFT is difficult to use for system descriptions mainly composed of dispersion (van der Waals) forces, such as gaseous systems, or systems in which dispersion contributes significantly, such as biomolecular systems [227]. Thus, numerous research studies examined the incorporation of van der Waals [228,229,230] to improve performance and enhance this technique. In addition to these constraints, the description of global potential energy surfaces of charge exchange excitations [230] is a prime restriction of DFT use in computational drug design. DFT usage overly favors sophisticated users and requires thorough reviews to determine the level of theory/methods to use for a particular system.