Applied Computer-Aided Drug Design: Models and Methods

Ligand and Structure-Based Drug Design (LBDD and SBDD): Promising Approaches to Discover New Drugs

Igor José dos Santos Nascimento¹, ², ³, *, Ricardo Olimpio de Moura³

¹ Pharmacy Department, Estácio of Alagoas College, Maceió-AL, Brazil

² Pharmacy Department, Cesmac University Center, Maceió-AL, Brazil

³ Programa de Pós-Graduação em Ciências Farmacêuticas (PPGCF), Departamento de Farmácia, Universidade Estadual da Paraíba, Campina Grande-PB, Brazil

Abstract

The drug discovery and development process are challenging and have undergone many changes over the last few years. Academic researchers and pharmaceutical companies invest thousands of dollars a year to search for drugs capable of improving and increasing people's life quality. This is an expensive, time-consuming, and multifaceted process requiring the integration of several fields of knowledge. For many years, the search for new drugs was focused on Target-Based Drug Design methods, identifying natural compounds or through empirical synthesis. However, with the improvement of molecular modeling techniques and the growth of computer science, Computer-Aided Drug Design (CADD) emerges as a promising alternative. Since the 1970s, its main approaches, Structure-Based Drug Design (SBDD) and Ligand-Based Drug Design (LBDD), have been responsible for discovering and designing several revolutionary drugs and promising lead and hit compounds. Based on this information, it is clear that these methods are essential in drug design campaigns. Finally, this chapter will explore approaches used in drug design, from the past to the present, from classical methods such as bioisosterism, molecular simplification, and hybridization, to computational methods such as docking, molecular dynamics (MD) simulations, and virtual screenings, and how these methods have been vital to the identification and design of promising drugs or compounds. Finally, we hope that this chapter guides researchers worldwide in rational drug design methods in which readers will learn about approaches and choose the one that best fits their research.

Keywords: CADD, Computational methods, Drug design, Drug discovery, Drug Development, Docking, FBDD, LBDD, QSAR, Rational Design, SBDD.

* Corresponding author Igor José dos Santos Nascimento: Pharmacy Department, Estácio of Alagoas College, Maceió-AL, Brazil; Cesmac University Center, Pharmacy Department, Maceió-AL, Brazil; and Programa de Pós-Graduação em Ciências Farmacêuticas (PPGCF), Departamento de Farmácia, Universidade Estadual da Paraíba, Campina Grande-PB, Brazil; Tel.: (+55)8299933-5457; E-mail: [email protected] & [email protected]

INTRODUCTION

The process of designing and developing new drugs is challenging and has evolved constantly in recent years, from empirical approaches related to natural products to the current phase with the use of computers and artificial intelligence [1-3]. One of the most significant advances in this area has been high-throughput screening (HTS), in which thousands of compounds can be screened in a few hours. In addition, the growth of genomics, proteomics, metabolomics, and molecular modeling promoted substantial advances in the knowledge of critical biochemical pathways for the P&D of drugs [4-6]. Associated with this, the synthetic approach exploring combinatorial chemistry could masterfully explore the available chemical space, supporting the discovery of new molecules [5]. However, the high financial cost and time-related to these approaches have driven researchers to adopt in silico methods [7, 8]. In this way, Computer-Aided Drug Design (CADD) emerged and perfected itself, indispensable in any new drug design discovery program [7, 9].

Traditionally, the discovery of a new drug can take between 10 and 15 years, with an investment of approximately US$800 million to US$1.8 billion [10, 11]. In this context, developing new drug design tools has become a constant quest to overcome old paradigms and speed up the discovery process at a lower financial investment [10, 12]. Over time, the scientific community accepted the new paradigm in the rational design of new drugs through CADD [13, 14]. The main reason is constant failures in the clinical evolution of prototypes identified and designed through classical techniques [13]. Thus, this paradigm shift facilitated the identification of new drugs, designing drugs with optimal physicochemical properties, and evaluating their potential in silico before they were synthesized [13]. With this, virtual screenings (VS) are increasingly explored, finding drug candidates in libraries of thousands of compounds. In addition, this method can be used in scaffolds identification as a starting point in molecular modeling studies, further confirming the in silico methods and rational design in the new era of P&D of drugs [15].

CADD can usually be divided into Structure-Based Drug Design (SBDD), and Ligand-Based Drug Design (LBDD) approaches. The researcher's choice between these approaches is related to the availability of key information about the clinical condition or known compounds against the same [16, 17]. For an SBDD protocol, the main requirement is the knowledge and availability of the target related to the explored clinical condition, in which the ligands are designed to interact with the target in question [18, 19]. On the other hand, in LBDD, there is no information about the target, but there are ligands of known activity against the clinical condition in question, and new molecules can be designed based on the production of pharmacophoric models or Quantitative Structure-Activity Relationship studies (QSAR) [20]. Traditionally, SBDD is preferred by the scientific community mainly due to the easy access to software and the wide availability of experimental structures of biological targets [21-23].

Currently, CADD methods using SBDD or LBDD approaches are vital in discovering new molecules and identifying critical information in drug design. Thus, this chapter will present a historical perspective on the evolution of drug design methods to CADD approaches. We hope that this chapter will guide drug developers in deciding on the type of strategy in their studies, increasingly promoting scientific advances in rational drug design.

DRUG DESIGN AND DISCOVERY: PAST AND TODAY METHODS AND OTHER APPROACHES

Strategies used in drug design and discovery have improved over the years [24]. In a historical context, each strategy was responsible for numerous discoveries. However, the improvement of methods made the process faster and more effective in the search for innovative molecules until the arrival of computational methods [25]. The following topics will address the evolution of the methods and their historical importance.

Natural Compounds (NC)

Before any study of rational drug design, Natural Compounds (NC) were the primary sources of drugs explored. During the last five decades, NCs were the target of isolation or total syntheses, as they presented high biological potential and challenging structural complexity. The discovery of numerous NCs against threatening diseases like cancer and infectious diseases has increased the interest in discovering new revolutionary NCs [26]. Indeed, most drugs introduced into the pharmaceutical market since 1994 are NCs or modified synthetic analogs, highlighting their potential for many years [27].

Traditionally, drug discovery by NCs starts with testing the extract of interest in vitro or in vivo assays. After demonstrating the pharmacological effect, the responsible compounds are then isolated [27, 28]. These compounds can be modified from then on to improve their pharmacological effect [27]. In a more current approach, drug repurposing using known NCs is used to find new potentials for available structures [29, 30]. Examples of natural compounds include Artemisinin (1), Atropine (2), Metformin (3), and Quinine (4) (Fig. 1). It is essential to highlight that these molecules were useful as molecular scaffolds that led to important clinical discoveries, which highlights the role of NCs in the R&D process [31]. However, synthetic difficulties, low quantities isolation, and challenges in optimizing these NCs led to the disuse of this approach. Thus, pharmaceutical companies and academics opted for rational strategies that promoted faster results with less financial costs and improved drug design techniques [27, 32].

Fig. (1))

Chemical structure of any drugs discovered by NCs.

Synthetic Drugs: Classical Approaches

The passing of the years revealed a decrease in the search for new drugs based on natural products. As previously mentioned, the low yield of isolated products, difficulties in purification processes, and toxicity related to the wild structural framework has increasingly instigated the synthetic approach to obtain new drugs. Furthermore, rational molecular modification strategies were proposed and improved over time, leading to the rational design era [33, 34]. These approaches will be presented in the following topics.

Bioisosterism

After identifying an NC with promising activity but an inadequate pharmaco- kinetic and/or toxicological profile, the need arose for strategies that could be coupled to organic synthesis aimed at a rational modification of this chemical scaffold [35]. In this sense, bioisosterism emerges as one of the most effective strategies available to medicinal chemists [36]. The requirements for the application of this strategy are i) knowledge about pharmacophoric groups; ii) mechanism of action; iii) metabolic inactivation pathways; iv) physical-chemical properties that determine its bioavailability and side effects [37, 38]. Pharmaceutical companies frequently use this strategy for sales competitiveness (me-too drugs) [38]. This strategy suggests changes that aim to modulate its pharmacological potency, physicochemical properties, and pharmacokinetics [39]. The success of this strategy can be exemplified by several marketed drugs, such as Piroxicam (5), Celecoxib (6), Zidovudine (7), and Cimetidine (8) (Fig. 2) through different bioisosteric replacement approaches [38].

Fig. (2))

Drugs designed using bioisosterism approach.

Molecular Simplification

In fact, part of the NCs present high structural complexity, which some authors call molecular obesity. This is one factor that makes using these structures unfeasible as drugs, as they have an inappropriate molecular pattern and are related to several side effects. For example, molecular simplification was applied to the Morphine (9) to obtain the Tramadole (12) (Fig. 3), showing less side effects and highlighting the promising potential if the molecular simplification approach [40]. In this context, the molecular simplification strategy arises, in which molecular groups not essential for the compound's activity are removed, representing one of the most commonly used strategies in lead optimization [41, 42]. One of the critical factors in this approach is the elimination of redundant chiral centers, reduction in the number of rings, and scaffold hopping. Thus, structural units are rationally removed to determine their biological importance, and based on structure-activity relationships (SAR) and pharmacophore groups, the essential units are maintained [41]. Furthermore, the reduction in molecular weight is considered a critical factor in improving the pharmacodynamic and pharmacokinetic profile of the simplified analog. Thus, simplifying the molecular structure while maintaining its pharmacological activity is also related to better synthetic accessibility of the analog and accelerates the discovery process. Classic examples of molecular simplification include Morphine (9) analogs such as Butorphanol (10), Pentazocine (11), and Tramadol (12) (Fig. 3) [40].

Molecular Hybridization

Another crucial molecular modification strategy performed through organic synthesis is molecular hybridization [43]. Through this, new ligands are designed based on the molecular recognition of two or more pharmacophoric subunits of compounds with known activity. Thus, the fusion of these subunits leads to a hybrid molecule that maintains the biological properties of the original templates, such as physical-chemical, pharmacological, and toxicological properties [44, 45]. Among the advantages of applying this strategy and producing multiple ligands, including i) the same molecule interacting with different targets, improving the therapeutic potential, and ii) improving bioavailability, and can be used in the production of prodrugs [46]. This strategy is exciting and constantly applied in developing drugs against clinical conditions involving multiple pathways, such cancer and Alzheimer Disease, in which a multitarget inhibitor would be desired [47-49]. Based on this strategy, Abourehab et al. performed the synthesis of ibuprofen and ketoprofen with pyrrolizine/indolizine aiming at anticancer activity. Thus compounds (13), (14), and (15) (Fig. 4) showed promising results against MCF-7 cells (IC⁵⁰ of 7.61, 1.07, and 3.16 µM, respectively). This study exemplifies the potential of this strategy, especially against multifactorial diseases such as cancer [50].

Fig. (3))

Morphine and their analogs discovered through molecular simplification.

Combinatorial Chemistry

The improvement of organic synthesis methods linked to the need to find new promising molecules resulted from combinatorial chemistry. One of its objectives is to carry out the serial synthesis of a large number of molecules in less time and financial cost and consequent biological evaluation of this structural diversity [51]. In addition to a synthetic approach, combinatorial chemistry can also be applied to NCs, where ample chemical space is explored to find promising hits and leads [52].

Fig. (4))

Drugs designed using a molecular hybridization approach.

Basically, in the combinatorial chemistry approach, a library of structurally diverse compounds is built through repetitive, systematic, or covalent linkages of building blocks (chemical groups). After their synthesis, they are tested against the biological targets of interest. The first studies with the application of combinatorial chemistry were carried out around the 1980s, in which several combinations of peptides synthesized through the solid phase were evaluated. Over time, this method was used for lead compound discovery and optimization within a combinatorial library [53].

Despite being an exciting and well-explored strategy, around the 2000s, academics and pharmaceutical companies stopped using this strategy mainly due to the delay in obtaining large libraries and failures in the clinical evolution of the compounds. It became clear then that synthesizing 1,000 to 10,000 compounds would not increase the probability of finding a promising candidate for pre-clinical trials, as discovering a drug is planning, not just numerical combinations. In addition, it is challenging to obtain 10,000 different varieties of a single chemical scaffold, as many times it could present a small structural diversity [54].

High Throughput Screening (HTS)

With the popularization of combinatorial chemistry, the need arose for technologies that could evaluate these large libraries of compounds quickly. Furthermore, the availability of multiple molecular targets has raised the need for technologies that can evaluate libraries against a wide range of targets. In this way, the High Throughput Screening (HTS) technology emerged, in which thousands of compounds can be evaluated in enzymatic or cellular assays, resulting in the discovery of useful leads and hits in drug design [55, 56]. Through HTS, around 10,000 to 100,000 compounds are screened daily, whereas over 100,000 can be performed by ultra-high throughput screening (uHTS) [56].

The most common readings on HTS are fluorescence and bioluminescence. Several fluorescence methods have been developed for HTS, the main ones being fluorescence resonance energy transfer (FRET) and fluorescence quenching energy transfer (QFRET). For example, the b-lactamase gene reporter assay is a widely used FRET-based assay. In addition, luminescence assays can be used, which present a lower signal than fluorescence, but with a more excellent range. The most used luminescence technology is luciferase reporter genes. Furthermore, detection in HTS can also be performed using atomic absorption spectroscopy, high throughput electrophysiology, absorbance, and scintillation proximity assays [57].

However, this approach fell into disuse mainly due to its high financial cost and the low quality of the data generated. The greater the number of compounds tested, the lower the quality and credibility of the results. Another problem is that the compounds are tested at a concentration, which can generate several false negatives [58, 59]. In addition, detection technologies are susceptible to false positives and negatives due to the physical properties of the compounds, which greatly limited and made researchers opt for more effective assay methods, driving the growth of pharmacological approaches in drug discovery [57, 60].

Target-Based Drug Discovery (TBDD)

The Target-Based Drug Discovery (TBDD) approach is one of the most traditional in drug discovery. This is an experimental approach in which compounds are screened against a specific biochemical target of the studied disease. A target can be conceptualized as a gene, gene product, or molecular mechanism identified through biological observations. A genetic target is a gene product or a gene that carries mutations that increase the predisposition to develop the disease (Ex. Alzheimer's disease, schizophrenia, or depression). On the other hand, a mechanistic target is an enzyme or receptor identified through biological observations or mechanism of action of known drugs [61, 62]. In this way, the assay against the specific target can be used to discover or design new compounds, identified as hits and leads, that can be optimized to improve their pharmacological potency [63, 64]. For many years, this strategy was one of the most used in drug discovery, and it resulted in the discovery of commercial drugs, such as Gefitinib (16), Imatinib (17), Raltegravir (18), and Zanamivir (19) (Fig. 5) [65]. However, using target-based approaches, which do not reflect the reality of the system, is a significant limitation of this technique. Furthermore, the target-based strategy based on a single target cannot simulate entire organisms, which can lead to later failures. Finally, the wrong choice of target can also lead to the identification of molecules that are not useful against the given clinical condition, which the Phenotypic-Based Drug Discovery (PBDD) approach tries to overcome this challenge [66].

Fig. (5))

Chemical structure of some drugs identified using TBDD approach.

Phenotypic-Based Drug Discovery (PBDD)

Unlike TBDD, in the Phenotypic-Based Drug Discovery (PBDD) approach, drugs are screened against a biological system, usually, cellular assays, in which the evaluation is based on phenotypic modification [67]. Also known as classical pharmacology or direct pharmacology, PBDD evaluates the phenotypic change free of any target hypothesis, in which several biochemical pathways are evaluated and may be involved in the activity of the compounds [68]. For example, inflammation or cancer involves multiples events, and in phenotypic screenings, several drugs targets can be involved in the compounds activity. This approach is historically more promising and related to several successful cases in drug discovery, such as Ezetimibe (20), Vorinostat (21), and Linezolid (22) (Fig. 6). However, not long ago, due to the higher cost and difficulties in optimizing the compounds due to the lack of knowledge of the molecular target, decreased its prevalence in research, and researchers began to opt for TBDD [65, 69]. But because it is a complete approach and uses more complete organisms to screen molecules, academics and pharmaceutical companies are returning to using PBDD as the main approach in their drug discovery programs [67, 70].

Fig. (6))

Chemical structure of some drugs identified using PBDD approach.

Multitarget Drug Design (MDD)

As shown in this chapter, drug discovery has been influenced for many years by many biological targets available. However, despite numerous advances in drug discovery, the number of successful drugs has been falling each year [71, 72]. It is increasingly evident that one of the reasons is the focus on just one target in the search for new molecules. Currently, a new concept of how drugs interact in the body is emerging, increasingly leaving the lock and critical model aside. This is mainly due to polypharmacology, side effects, and drug repositioning. Thus, the concept that a drug interacts with multiple targets grows more and more, in which academics are trying to find a drug and its various biological targets [73, 74].

In the mid-2000s, the general principles of Multitarget Drug Design (MDD) were proposed. Thus, a lead compound that interacts with multiple targets can be a single molecule or composed by the fusion of several structural nuclei in which each one interacts with a different target. One of the requirements is that these fused cores are smaller than the molecule that binds directly to the target, and their separate core binding efficiency is lower than that of clinically used ligands. After identifying the potential drug, it can become an optimized lead to improve activity, selectivity, and physicochemical and pharmacokinetic properties. Finally, multitarget drug design constantly uses in silico methods [75]. Advances in this technique are evidenced by Entrectinib (23) (Fig. 7), a multitarget inhibitor that acts against the MAPK, PLC-γ, and PI3K pathways in the treatment of nonsmall cell lung cancer with a positive ROS1 fusion gene. Another example is Imeglimin (24) (Fig. 7), an antidiabetic that acts inhibition of the hepatic production of glucose and amplification glucose-stimulated insulin secretion (GSIS) [49].

Computer-Aided Drug Design (CADD)

As shown so far, designing and discovering new drugs is time-consuming and demands a high financial cost. In this context, strategies that could predict the potential of molecules before synthesis and biological evaluation would be ideal since they would make the process agile with less investment [76]. Thus, in silico methods of Computer-Aided Drug Design (CADD) appear, in which computer simulations are used to predict the pharmacological potential of a given ligand, which is extremely useful in the initial processes of any drug development campaign [76]. Technological advancement and the Big Data era benefited academics and pharmaceutical companies from these methods. These include the availability of large databases of biological targets and ligands and broad access to high-performance computing and high-throughput software, making essential CADD methods in searching for new drugs [77, 78].

Fig. (7))

Approved drug with Multitarget mechanism.

CADD is a multidisciplinary tool used as a shortcut to discover, analyze and develop new drugs. This tool can identify targets and ligands or promote rational changes in lead compounds through two main approaches: SBDD and LBDD (discussed in the next topic). Furthermore, this tool is constantly used to optimize the pharmacokinetic properties of ligands related to absorption, distribution, metabolism, excretion, and toxicity (ADMET). Furthermore, essential require- ments for using these tools are practiced in a computational environment to speed up the lead identification/optimization process. As an advantage, it can remove molecules with unwanted properties, choosing only the most promising ones [79]. The success of this technique is highlighted mainly in the discovery of the Human Immunodeficiency Virus (HIV) protease inhibitors Saquinavir (25), Indinavir (26), Ritonavir (27), Amprenavir (28) (Fig. 8), and others that were revolutionaries in the treatment against this virus and continue to be explored in the search for more potent inhibitors [80].

SBDD AND LBDD METHODS IN DRUG DESIGN

Structure-Based Drug Design (SBDD)

A classical protocol of SBDD is applied when there is experimental information about the drug target, with their 3D co-crystallized structure complexed with a ligand. In this way, millions of molecules can be screened and chosen based on the best affinity and complementarity with the target binding site [20, 81]. Molecular docking is a powerful tool used in these protocols, generating information about the interactions between ligands and macromolecules and the best conformation to binding in the active or binding site. In fact, this tool used with pharmacophore models can lead to a more specific screening of ligands [81]. On the other hand, if the target is known and the experimental 3D structure is unavailable, SBDD protocols can be applied using homology modeling. In this way, the target structure is built based on a homologous structure [81]. The following topics will show the main approaches used in SBDD campaigns.

Fig. (8))

Chemical structure of some HIV protease inhibitors discovered by CADD methods.

Homology Modeling

For success in an SBDD protocol, knowledge about the structure and function of the biological target is essential [82]. Thousands of experimental structures of these drug targets can be easily found in the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB). In addition, thousands of protein and amino acid sequences are available in the National Center for Biotechnology Information (NCBI) database, which can be used for alignment and homology verification or even model building. In fact, the number of sequences is greater than that of experimental 3D structures, which makes homology modeling a powerful tool in drug design by SBDD [83]. As the number of sequences is greater than the number of experimental structures, computational methods are constantly used for the structure prediction of these sequences. Through homology modeling, a 3D model of a target protein is constructed using an amino acid sequence (usually in FASTA format), generating a 3D structure for the intended target, which has not been characterized experimentally [84, 85].

A similarity above 30% between the experimental structure and the sequence is necessary for the targets to be considered homologous. Basically, a procedure for building a model by homology can be applied as follows: i) identification of the 3D structure homologates the sequence; ii) Alignment of sequences of the 3D structure and the homologous sequence; iii) construction of the model from the alignment; iv) validation of the constructed structure. This procedure can be repeated with other sequences if the model is not adequately validated [86]. A powerful tool used in this type of protocol is the SWISS-MODEL web server, which performs all these steps and is widely used by the scientific community [87].

Homology modeling was helpful during the COVID-19 pandemic due to the pathological agent SARS-CoV-2 since it was necessary to identify a drug that could prevent this disease from spreading faster and faster worldwide. As an example, one can cite the study by Zhang and Zhou (2020) [88] in the search for a promising compound against RNA-dependent RNA polymerase (RdRp) that could be used against this disease. Thus, RdRP was constructed through homology modeling, using the SARS-CoV RdRp (Identity of 95.8%) as a template. Then, molecular dynamics simulations were performed to obtain the native structure of the protein. Finally, the authors identified Remdesivir (29) (Fig. 9) as a potential inhibitor of RdRp. It should be noted that this drug was approved for hospital use in patients infected with SARS-CoV. This study highlights the importance of homology modeling in the drug discovery process.

Fig. (9))

Chemical structure of Remdesivir.

Molecular Docking and Molecular Dynamics Simulations

Molecular docking and Molecular Dynamics (MD) simulations are essential tools for any drug discovery and development campaign. Through these tools, it is possible to rationally design new compounds and model biochemical processes to propose mechanisms of action and interaction involved in the target inhibition process. Thus, target conformation and ligands favorable to the inhibition process are revealed, helping researchers to search for innovative molecules. This is an in silico identification, which can be performed on large ligands libraries even before the compound synthesis [89, 90].

Molecular docking aims to identify the best ligand conformation in its biological receptor (binding or active site). In this way, several conformations, known as binding modes or poses, are generated, and those with greater affinity for the region are analyzed. Clearly, the target 3D structure is essential for this approach. If it is unavailable, homology modeling can be used for structure construction (as explained in the previous topic). Based on this information, this technique is based on: i) Generating the binding poses and ii) Ranking these conformations and choosing the one with the most excellent affinity for the interaction site [91].

One of the problems of molecular docking is that current programs do not consider target flexibility, which can lead to a mistake in identifying promising molecules. These problems can be solved using MD simulations, assuming the flexibility of proteins and ligands. In addition, more modern methods such as Molecular Mechanics/Poisson-Boltzmann and Surface Area (MM-PBSA) are constantly used to estimate the binding energies of a ligand for its target with values closer to the real, increasing the possibilities of identifying active molecules [92].

With advances in drug design, it is observed that the key-lock model that explains how a drug interacts with its biological receptor has fallen into disuse. This occurs mainly because this model does not consider the proteins' flexibility in interacting with the ligand. Thus, several studies of molecular docking highlight that it is essential to predict the conformation of the receptor during the interaction process. This decreases the probability of identifying false positives in drug discovery campaigns. Thus, a couple of MD simulations and Molecular Docking are constantly used. MD simulation is used to find the best match of a target before the docking protocol or even the use of docking to obtain the complex, followed by an MD simulation to identify the most appropriate target and ligand conformation [93].

The combination of MD simulation protocols with molecular docking is responsible for identifying several commercial drugs. One of the most current is Vaborbactam (30) (Fig. 10), launched on the market in 2017. It is one of the newest β-lactamase inhibitors, useful as an adjuvant in antimicrobial therapy. Its genesis was based on the flexible docking strategy through the ICMdocking software [94]. Other marketed drugs discovered through docking and MD simulations include the protease inhibitors Amprenavir (28) (Fig. 8), approved for clinical use in 1999, in which MD simulations proposed a weak hydrogen bond of an amide group, replaced by N,N-dialkyl sulfonamide, improving affinity for the enzyme and resulting in the highlighted compound. However, Amprenavir (28) was discontinued in 2004 and replaced by its prodrug, Fosamprenavir (31) (Fig. 10) [94]. Another drug discovered by coupling docking with MD simulations is Raltegravir (18) (Fig. 5) [80].

Fig. (10))

Chemical structure of the drugs discovered using MD simulations and molecular docking.

Fragment-Based Drug Design (FBDD) or de novo Drug Design

Another strategy widely explored in drug discovery campaigns is the Fragment-Based Drug Discovery (FBDD) strategy, which aims to find active molecular fragments and optimize them to improve activity against a specific target [95, 96]. One of the first authors who presented this strategy was Fesik and collaborators (Abbott Laboratories) in 1996, with its conceptualization carried out by Hol and colleagues in the 90s [95]. Since then, FBDD has been associated with several successful cases of discovering new drugs where other methods failed to achieve the same objectives [95, 96].

The first step in an FBDD protocol is constructing a library of low molecular weight molecular fragments designed against a specific target [97]. These fragments must have sufficient size for the interaction site and low chemical complexity so that the interaction occurs most favorably, avoiding unfavorable interactions [96, 97]. Thus, the initial focus is the identification of attractive fragments, followed by their optimization. A fragment must have a weak binding affinity in the range of mM and μM [96]. Thus, the optimization of the power of this fragment is obtained through the fusion, linking, and growing strategies, better-known approaches in FBDD campaigns [96, 97].

The growing approach is preferred by researchers using FBDD [95]. In this approach, a fragment is placed at the target interaction site to perform significant interactions, and this fragment grows to complement this binding site [95]. This protocol is carried out on an experimental structure of the target, obtained by X-rays or NMR, with the placement of the fragments through molecular docking [95, 96]. During the process, it is essential to maintain the interactions of the initial fragment in the optimized compost to obtain a molecule with superior activity [95].

Although less used than the growth approach, the ligation approach is quite versatile in designing promising molecules and optimizing fragments [95]. Thus, the basis of this approach is the addition of fragments in strategic locations of the target binding site, followed by the addition of linkers to unite them and form a single compound with optimized properties [95, 98]. Although simple, this approach is more challenging, as it requires optimal linkage between fragments [98].

On the other hand, the de novo drug discovery strategy, similar to previous approaches, is based on the 3D structure of the receptor and pharmacophore groups that constitute the ligand [99]. This approach uses optimizable lead compounds from zero to design new chemical structures adapted to the selected target binding site [100]. There are two main ways of using this approach: i) outside-in, also known as linking, in which fragments are added to interact at the main sites of the binding site, followed by linking these fragments to obtain a single compound; and ii) inside-out, known as growing, in which a fragment is added at the binding site, and it grows in such a way as to present complementarity and essential interactions with the interaction site resulting in the formation of an optimized compound [100].

It is possible to observe that the concepts of FBDD and de novo drug discovery are similar. Interestingly, the differences between the two approaches are little explored in the literature, and in most cases, the concepts are confused. Some authors point to the building blocks as one of the differences between the two approaches. In de novo drug design starts with smaller fragments. However, the fragment size is not precise, and thus the definition between de novo or FBDD is defined by the study authors due to this literature gap [101].

As shown in FBDD and other techniques presented in this chapter, new drug design is explored through computer-aided drug design. Caveat and SPROUT are software exploited using this technique, which places fragments using an outside-in method. On the other hand, the inside-out approach can be applied through the LUDI software [100]. In addition, the LigBuilder V3 program has become a fundamental tool in drug design through the de novo approach [102].

The use of this technique resulted in the discovery of Dacomitinib (32) (Fig. 11) by researchers from Pfizer performed. Thus, in vitro screening of molecular fragments was performed. The compound was built using different combinations to synthesize new analogs against the epidermal growth factor receptor (EGFR) in the search for new anticancer drugs. Finally, Dacomitinib (32) was chosen because it had better pharmacokinetic properties [103]. On the other hand, using virtual fragment screening in a collaboration between academics and pharmaceutical companies (Janssen), Erdafitinib (33) (Fig. 10) was discovered, an inhibitor of Fibroblast Growth Factor Receptors (FGFR) in cancer cell lines (IC50 between 0.1 to 130 nM) and selectivity against the inhibition of Vascular Endothelial Growth Factor Receptor 2 (VEGFR2). Its potency and excellent pharmacokinetic properties resulted in its approval in clinical trials and later approval by the FDA Food and Drug Administration [104]. Finally, Zanubrutinib (34) (Fig. 11) was discovered through FBDD as a Novel Covalent Inhibitor of Bruton's Tyrosine Kinase, useful as an anticancer drug [105].

Fig. (11))

Chemical structure of the drugs discovered through FBDD or de novo drug design.

Density Function Theory (DFT)

As shown so far, SBDD is based on identifying promising molecules based on their binding energies. However, more accurate methods are still needed to calculate these energies and the interactions of a ligand with the macromolecule. In this context, there is a growing tendency to use quantum mechanics (QM) through CADD approaches. This is due to advances in computing and trends in QM providing more accurate data, making it a powerful tool for medicinal chemists [106, 107]. One of the most explored methods in QM is the Density Functional Theory (DFT), in which ground state energy and other molecular properties are determined through their electronic densities and how they are related to the pharmacological activity of a molecule [108, 109].

Currently, DFT is widely used in the design of drugs in active sites of enzymes, inhibiting their catalytic activity. Despite molecular docking not being a handy tool in these studies, DFT can calculate affinity energy values more accurately, increasing the predictive ability to find new drugs [110]. It is used to predict the most significant interaction forces, yielding critical information that can be used to design new drugs [108]. One of the most explored applications is predicting covalent bonds between an inhibitor and an enzyme. Through this method, it is possible to predict the binding energy for the formation of this type of bond and predict whether the drug can act covalently, an approach that has been widely explored in the design of anticancer, antimicrobial, and antiparasitic drugs [108, 111].

QM also provides ab initio methods, widely used in drug design, which require less computational power when compared to DFT. Similar to DFT, they use the electronic Schrödinger equation to calculate binding energies, electron densities, atomic and nuclear coordinates, and other parameters. Thus, using this approach, it was possible to discover Dorzolamide (35) (Fig. 12), a carbonic anhydrase (CA) II inhibitor used in the treatment of