Applications of Molecular Docking Studies in SARS-CoV-2 Targeted Drug Discovery and the Gains Achieved through Molecular Docking

Abstract

In this chapter, we delve into the pivotal role of molecular docking in the realm of computational biology and chemistry, focusing specifically on its application in drug discovery targeting SARS-CoV-2. Molecular docking, a critical computational technique, has played a significant role in predicting the interactions and bindings of molecules, particularly concerning SARS-CoV-2’s main protease and RNA polymerase. This chapter highlights the synergy between molecular docking and virtual screening, emphasizing the expedited identification and evaluation of potential drug candidates against SARS-CoV-2. Through a comprehensive discussion, we aim to provide a nuanced understanding of the rapid advancements in drug discovery for SARS-CoV-2, accentuating the indispensable value of computational tools and methods in contemporary therapeutic development.

1. Introduction

Severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV) are viruses in the family Coronaviridae that infect different animals and can cause mild or severe respiratory infections in humans. These viruses were detected in humans in 2002 and 2012, with fatal outcomes. In late 2019, a new coronavirus emerged in Wuhan, China. The International Committee on Taxonomy of Viruses named it SARS-CoV-2, while the WHO named the disease COVID-19. This virus multiplied rapidly in a short time and spread all over the world [1]. As of October 14, 2023, the World Health Organization reported a cumulative total of 774,291,287 COVID-19 cases and 7,019,704 deaths worldwide. Vaccination began on June 22, 2020, and a total of 13.59 billion vaccine doses have been administered “https://.covid19.who.int/ (accessed on 29 January 2024)”.

Coronaviruses are enveloped, positive-sense single-stranded RNA viruses with genome sizes ranging between 26 and 32 kilobases. The genomic structure comprises various components that form the enveloped virion. The genome of SARS-CoV-2 shows approximately 80% similarity to that of SARS-CoV. The virion consists of genomic RNA, phosphorylated nucleocapsid (N) protein, membrane (M) protein, envelope (E) protein, and spike glycoprotein (S) as fundamental building blocks (Figure 1). The life cycle of SARS-CoV-2 begins with the binding of the viral protein S to the angiotensin-converting enzyme-2 (ACE2) receptor [2, 3, 4, 5, 6].

Figure 1.

The schematic structure of the SARS-CoV-2 virus created in BioRender.com (accessed on January 31, 2024).

Although the pathogenesis of SARS-CoV-2 infection remains uncertain, severe COVID-19 is characterized by an uncontrolled immune response induced by the host’s reaction to the virus. This leads to a sudden and explosive release of various immune mediators, particularly cytokines, and damage-associated molecular patterns (DAMPs). The unregulated immune response can result in multiple organ dysfunction, leading to sepsis or septic shock, and ultimately giving rise to a life-threatening syndrome [4]. Numerous studies have been conducted on viruses within the Coronaviridae family. It has been revealed that the ACE-2 receptor plays a crucial role in the virus’s entry into the host. Additionally, RNA-dependent RNA polymerase (RdRp), main protease (Mpro), helicase, papain-like protease, and mTOR signaling pathway [7] in the virus are identified as potential targets [8].

The COVID-19 virus can cause a range of symptoms, including a dry cough, high fever, body pain, and shortness of breath. These symptoms may appear within 2–14 days of exposure to the virus. In more severe cases, patients may develop pneumonia, as well as neurological symptoms such as headache, loss of taste and smell, and visual impairment. It is important to seek medical attention if you experience any of these symptoms [6, 9].

Since the commencement of the pandemic in 2019, the virus has undergone numerous changes, giving rise to the development of novel variants and their subvariants. Each variant carries unique genetic markers that may impact its transmissibility, virulence, and ability to evade the body’s immune system [10]. The fact that there are many different variants of the virus, in addition to the well-known ones such as Alpha, Beta, Gamma, Omicron, and Pirola, highlights the urgent situation we are facing. This is evident from the high rates of transmission and mortality. As a result, it is crucial to identify effective drug targets, antiviral drug molecules, and strategies to contain the spread of the virus quickly and accurately [11].

Due to the limited number of drugs approved to treat COVID-19, it has become a significant threat to global health, safety, and the economy. The urgency to find antiviral agents to combat the spread of the SARS-CoV virus has made it necessary to expedite drug discovery and development. Researchers have had to employ new strategies to hasten traditional processes and discover new drug candidates more quickly. One of the most crucial solutions has been the use of computer-aided drug design (CADD). CADD is a computational approach used to discover, develop, and analyze drugs and active molecules with similar biochemical properties. It can be applied to all stages of drug development, making it essential because of its high prediction accuracies and unique conveniences in terms of time and resources. As a result, the rapid advancements in drug discovery during this period increased the scientific community’s capacity to provide faster and more effective solutions to urgent health issues [12, 13, 14].

Once the target is determined, the process of selecting candidates for in vitro testing begins. This step is slowly progressing without the aid of computer-aided drug design, also known as in silico studies. However, by applying in silico studies, it is possible to select the compounds most suitable for the target structure and eliminate the unsuitable compounds. This approach helps speed up the process and achieve a higher success rate more cost-effectively. At the end of in vitro and in vivo experiments, a potential drug molecule is obtained (Figure 2) [15].

Figure 2.

The role of computer-aided drug design in the drug discovery process.

2. Molecular docking in computational biology

Molecular docking is a computational tool that plays a crucial role in computer-aided drug design. It is a molecular mechanic approach that aims to fit a ligand into a three-dimensional binding site. It helps in identifying potential drug candidates and studying the interactions between molecules at the molecular level, which is essential for drug design. This technique can predict the conformation of molecular interactions and calculate the binding energy scores of these interactions. It simulates the process of evaluating how efficiently a drug candidate can bind to a target molecule. Molecular docking has become a routine part of computer-aided drug design and plays a vital role in drug discovery [16, 17, 18].

With advances in the field of drug discovery, the use of in silico tools has become significantly widespread today. Molecular modeling contributes significantly to drug design processes, especially by providing a critical tool to understand the fundamental interactions of ligands and the mode of action of the target molecule. Methods such as molecular docking are effective in identifying different modes of docking through key docking scores and binding interactions. These in silico assessments contribute to the rapid and cost-effective identification and optimization of potential drug candidates, improving the efficiency of the drug discovery process [19].

Many research groups are exploring drug repurposing, focusing on in silico evaluations of approved drugs. Drug repurposing is a process that involves using existing drugs for new indications. Compared to traditional drug development processes, drug repurposing offers several advantages. Firstly, it facilitates a faster development timeline due to the already proven safety of drugs used in humans. By bypassing early clinical trial stages, the timeline is shortened, and costs are reduced. Secondly, utilizing approved drugs for new indications accelerates the creation of new treatments, decreasing the costs of drug discovery and development while enhancing the potential for rapid access to innovative therapies. However, there are challenges and limitations associated with drug repurposing [20, 21, 22]. Data from 2020 showed a significant increase in virtual screening studies to find new potent molecules or repurpose approved, withdrawn, or orphan drugs effective against the virus [13].

There are different types of software available to perform molecular docking, which include AutoDock [23], AutoDock Vina [24], DOCK, FlexX [25], Molecular Operating Environment (MOE) [26], and Glide [27, 28]. These software programs have been used in various molecular docking studies such as Flexx and AutoDock Vina for Mpro [29, 30], AutoDock for proteins [31, 32], MOE for ACE-2 [33], and Glide for Rbrd [34], which are the target structures of SARS-CoV-2, with promising results.

The process of molecular docking involves evaluating the quality of the pose using a scoring function, along with predicting the position and conformation of the ligand in the protein binding site. Another crucial task in docking is scoring active compounds higher than known inactive ones, but this can be difficult due to various factors. The sampling process has many degrees of freedom, such as the rotation and translation of one molecule relative to another. Therefore, efficiently sampling conformational space is still a challenge in molecular docking. Early approaches reduced the degrees of freedom by treating the ligand and protein as rigid bodies, but those relied on shape similarities between the ligand and the protein binding site.

Modeling molecular flexibility poses challenges in terms of computational efficiency, as conformation is linked to protein-ligand interactions. As a result, many docking programs now consider the entire conformational space of the ligand. Thus, molecular docking algorithms aim to sample the conformational space of the ligand and protein and evaluate their interactions. However, they can face challenges like flexibility, computational efficiency, and scoring accuracy [35].

Previous studies have identified three types of scoring functions for protein-ligand interactions: physics-based, empirical, and knowledge-based. However, these scoring functions have limitations, and no single method can perform perfectly in all aspects. To improve our understanding of scoring functions, a new classification scheme has been proposed. This new scheme categorizes scoring functions into four groups: physics-based, empirical, knowledge-based, and machine learning-based. Hybrid scoring functions can improve performance by combining different scoring functions. These scoring functions can compensate for the weaknesses of different scoring functions. However, the computation time of hybrid scoring functions may be longer.

Machine learning-based scoring functions can outperform traditional scoring functions and improve prediction accuracy by using extended training datasets. These scoring functions may be more widely used in the future. Combining different types of scoring functions and using new types of features can improve the performance of scoring functions in molecular docking. Additionally, these scoring functions are open source [36].

The optimization of the scoring function is done by using large protein-ligand complex structures that have been experimentally determined. In recent years, the development of technologies such as infrared spectroscopy, X-ray crystallography, nuclear magnetic resonance spectroscopy, and cryogenic electron microscopy has increased the possibilities for protein-ligand complex structure determination. This has led to efforts to optimize scoring functions and improve sampling procedures. One example is the Lamarck genetic algorithm used in AutoDock Vina, which is implemented in an OpenMP-based multithreaded version to search for ligand conformations. There are also efforts to improve the performance of these algorithms by using Graphics Processing Units (GPU)-based accelerations [37].

The COVID-19 pandemic has highlighted the importance of using computer simulations to screen small molecules for potential viral inhibitors. Despite the promising results of molecular docking studies, receptor flexibility, especially the movement of the backbone and other important secondary elements, can make molecular docking difficult. Nevertheless, successful applications have demonstrated the efficiency of computational approaches for screening large databases and designing new molecules. To further advance molecular docking, the development of accurate and cost-effective scoring functions is crucial. However, it is important to note that experimental technology still plays a critical role in producing realistic molecular interactions [16, 38, 39, 40]. After conducting molecular docking studies, experimental studies are carried out to further analyze the structural and spectral properties of the molecules. Through numerous studies, the computational analysis of various properties is compared to experimental findings. This process leads to the confirmation of the molecular docking data by comparing it to the experimental results [41, 42].

3. SARS-CoV-2 targeted drug discovery

Ongoing research is being conducted to develop antiviral agents that can be effective against coronaviruses. These studies have identified several potential drug targets, including the human receptor ACE2, which facilitates the entry of the virus into the host, the spike protein of the virus, RdRp, Mpro, and papain-like protease [8, 11, 43].

Although SARS-CoV-2 interacts with all drug targets, its interaction with the active site residues of the Mpro was found to be much stronger than with the other targets, spike and ACE2. On the other hand, the spike protein has been used in vaccine studies, but frequent mutations in this protein have been an obstacle. The mutation observed in the spike protein was not observed in the Mpro. These reasons have put the Mpro several steps ahead of other targets in drug discovery and use as a drug target [44, 45, 46].

There are approved drugs with proven efficacy on the Mpro and RdRp among the targeted structures. Mpro and RdRp 3D structures from the protein data bank [https://www.rcsb.org/ accessed on: 27.01.2024] were visualized in the UCSF Chimera v1.17.3 program (Figure 3). The virus requires the enzyme Mpro for proteolytic processing of polyproteins. The active site of Mpro is primarily characterized by a catalytic dyad composed of His41 and Cys145. These residues are crucial for the enzymatic activity, as they facilitate the hydrolysis of peptide bonds within the polyprotein. Inhibitors designed to target this active site aim to disrupt the proteolytic process, thereby impeding viral replication. Mpro inhibitors inhibit viral replication and transcription of the virus by binding to the active site of the protease (Figure 3A). Rupintrivir, Lopinavir/Ritonavir, Nirmatrelvir, and Paxlovid [nirmatrelvir + ritonavir] are examples of this group of drugs [2, 3, 8, 14, 39, 49, 50].

Figure 3.

A: 3D structure and protein-ligand interactions of SARS-CoV-2 main protease (Mpro) with Nirmatrelvir (PDB:8E25) B: RNA-dependent RNA polymerase (RdRp) with Remdesivir triphosphate (PDB: 7UO4) [47, 48].

RdRp is a critical viral enzyme that catalyzes RNA replication from an RNA template and mediates viral replication and transcription. The catalytic activity of RdRp involves the synthesis of complementary RNA chains for a specific RNA template, presenting potential therapeutic targets against viral infections, as this activity is not essential for the survival of eukaryotic cells [51]. The active site of the RNA synthesis process comprises cofactor Mg2+ and several motifs (A to G), with motifs A and C being particularly important. These motifs include residues Asp618, Asp760, Asp761, Asp762 (motif A), and Ser861, Lys545, and Arg555 (motif C), which play a central role in catalysis (Figure 3B). Nucleotide analogs designed to inhibit the replication process of the virus target the mechanism of nucleotide addition at this active site. Remdesivir and Molnupiravir in this group inhibit replication by stopping the development of the RNA chain [14, 52].

A significant number of new variants and subtypes of the virus have emerged rapidly, each carrying unique genetic signatures that may influence their contagiousness, virulence, and the potential to evade the immune system. Alongside this diversity, challenges in in silico studies include inadequate resolutions of crystallographic targets, both structural and conformational flexibility issues, as well as simplifications and assumptions that adversely affect the accuracy of scoring functions. Additionally, limitations in properly accounting for solvent and entropic effects, hydrogen bonding, directional interactions, etc., can lead to difficulties [10, 13].

Most molecular docking software programs rely on force field calculations driven by quantum mechanics [QM] and experimental data to predict binding energy. However, accurate binding energies can only be determined through methods such as density functional theory [DFT] and molecular dynamics simulations. Therefore, it is crucial to employ multiple methods, rather than relying solely on the in silico approach [13].

A study titled “An oral SARS-CoV-2 Mpro inhibitor clinical candidate for the treatment of COVID-19” was published in 2021. It discusses the discovery of the oral drug PF-07321332/ Nirmatrevir, which was developed as a Mpro inhibitor to treat COVID-19. The drug has reached Phase I stages and has been proven effective against all coronaviruses. The study provides information on the bonds formed between the target and the ligands, bond lengths, type, and degree of interaction. For example, the study showed that the number of hydrogen bond donors was linked with poor oral bioavailability. Therefore, nitriles and benzothiazol-2-yl ketones were considered. When the nitrile compound was used, there was an increase in oral absorption in rats, indicating that the interaction could be improved by taking advantage of the structural features. The study also considered the structural cavities of the protease. Introducing suitable compounds into these pockets creates hydrogen bonds and an interaction. The drug PF-07321332/Nirmatrevir has been identified as a potent inhibitor of SARS-CoV-2 Mpro biochemical activity with its nitrile group [52, 53].

We observe not only in this study but in many studies that target-ligand complexes achieve results by taking advantage of interactions. These findings highlight the importance of molecular docking in the drug discovery process. In a study examined against the Omicron variant, it was explained that different types of interactions between ligand atoms and protein residues determine the extent of inhibition of the target protein. In this research, in silico studies, molecular docking and molecular dynamics were used. The number of hydrogen bonds in the docking results of the compounds found to be active was considered the most important of these interactions, as it ensures target-ligand stability. Additionally, other types of interactions such as pi-alkyl, pi-cation, and pi-sigma were also observed in the active compounds. The importance of these interactions in the activity is emphasized [12, 54, 55].

4. Integration of molecular docking and virtual screening

Virtual screening, considered a computational method proven and appreciated, serves as a computer-aided complement to experimental high-throughput screening, which is a contemporary of hit identification and optimization. It can be classified into two main categories: ligand-based and structure-based methods. Ligand-based methods, such as pharmacophore modeling and quantitative structure-activity relationship (QSAR) techniques, can be employed when there is minimal or no structural information available for the targets, and a series of known active ligand molecules exists. On the other hand, molecular docking stands out as a widely applied method in structure-based drug design since the early 1980s. Approved in 1981, Captopril (Capoten) was developed using structure-activity relationship (SAR), structure-based drug design (SBDD), and ligand-based drug design (LBDD). The use of CADD has continued to increase since then [13, 56]. Various programs based on different algorithms have been developed to conduct molecular docking studies, contributing to its growing significance as a valuable tool in pharmaceutical research [40].

Cladribine, which targets the adenosine deaminase protein, is an active substance approved in 1993 with the integration of virtual screening and docking methods [13]. In drug repurposing studies, it has been observed that the combined use of two in silico methods results in higher accuracy compared to using each method separately [57].

5. Successes and ongoing challenges

CADD methods greatly speed up early drug discovery, allowing rational selection of drug-like compounds for therapeutic needs [58]. To ensure the success of the docking experiment, it is crucial to select the appropriate docking software and protocol [59]. Recently, platforms have been available that combine various virtual screening approaches, support parallel computing, and provide flexibility to users. By optimizing data processing, seamlessly integrating CADD software, and leveraging parallel computing architecture, we aim to enable users to complete data analytics projects faster and more efficiently [60].

According to research reports, Artificial Intelligence (AI) drug discovery firms are creating high expectations for research and development by reducing costs and accelerating timelines, once molecules are in clinical trials [61]. AI and machine learning are also used to understand the basis of diseases and evaluate the efficacy of drugs. In drug development, AI can be useful in analyzing molecular structures and identifying potential drug candidates. Predicting protein binding sites and performing structure-based virtual screens are crucial steps in the drug development process. These technologies can potentially speed up the process, reduce costs, and facilitate the discovery of more effective treatments. However, there are challenges in this area such as choosing the right dataset, model training, and result validation [62].

We can also talk about deep learning as another advancement. Deep Learning (DL) is a methodology that has been used for language and image processing for decades. However, its applications in drug discovery have emerged only in the past few years. This acceleration is attributed to the use of Graphics Processing Units [GPUs] to overcome computationally expensive calculations associated with deep learning. Compared to traditional machine learning methods, DL uses various processing layers known as neurons to make predictions based on large collections of multidimensional data. There are various types of deep learning architectures available, but prominent applications in drug design and discovery include Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM), and Multi-Task Learning (MTL). The creation of new drugs targeting previously unaddressed drug targets is anticipated to enhance patient care standards by increasing therapeutic efficacy and reducing side effects. With the increasing prevalence of newer techniques like deep learning in academic drug design research, a significant increase in the efficiency of conducted studies is expected [63].

Despite the significant advancements in computer-aided drug discovery, the time required to launch a new drug has not reduced significantly. Studies indicate that many active compounds found through computational techniques exhibit similar activity to compounds in clinical trials. However, most of these compounds cannot survive clinical trial stages due to the lack of necessary pharmacokinetic properties, despite demonstrating good antiviral activity and a well-defined mechanism of action. This highlights the importance of developing additional and complementary tools for evaluating pharmacokinetic properties and off-target effects. In silico approaches can provide pre-selection for candidates, while biological validation of computational predictions is also necessary. With the increasing availability of curated experimental datasets, physics-based methods and artificial intelligence techniques are expected to play a more supportive role in evaluating both the pharmacodynamic and pharmacokinetic properties of investigated compounds. Lastly, an increased probability of highly contagious disease outbreaks, underscoring the potentially significant role of successful computational strategies in combating future challenging diseases [64].

6. Conclusion

Computer-aided drug design (CADD) methods are a useful tool in the process of drug discovery. It is an area of technology that is currently developing and offers faster, more economical, and higher success rates in drug discovery. The most used CADD method is molecular docking. However, the use of multiple methods increases the success rate of the results by creating synergistic effects. The molecular docking examines the docking and interactions between the target structure and possible drug molecules. Many programs have been developed and continue to develop with technologies such as artificial intelligence, machine learning, and deep learning.

The COVID-19 pandemic has had a significant impact on the world since 2019. Although the virus is like other previously discovered members of the family, effective treatments must be developed. The rapid spread of the virus, the emergence of various variants, and the increasing number of cases and deaths have necessitated the acceleration of the drug discovery process. Multiple targets have been established in studies of the virus. Among these targets, Mpro and RdRp are two prominent proteases. These proteases are critical enzymes in virus proliferation, making it challenging to find effective drug molecules.

Drug discovery is a lengthy and exhaustive process that involves identifying new molecules that can be used to create a drug. One approach to accelerate this process is to reuse or repurpose molecules that have already been approved for other uses. This approach can help to shorten the process to some extent. Virtual screening, molecular docking, and molecular dynamics methods can be used to screen molecules in silico before further in vitro studies. This enables the interaction between molecules to be calculated on a computer.

This book chapter discusses the structural and pathological properties of the SARS-CoV-2 virus, which caused the COVID-19 pandemic. It also discusses the advantages and disadvantages of using molecular docking as a method called CADD. The chapter also covers drug discovery processes against the SARS-CoV-2 virus using molecular docking. It gives a case study and multiple examples of how molecular docking studies contribute to the prediction process and drug discovery processes. Finally, the chapter concludes with current and future expectations.

Conflict of interest

The authors declare no conflict of interest.

Appendices and nomenclature

SARS-CoV	Severe acute respiratory syndrome coronavirus
MERS-CoV	Middle East respiratory syndrome coronavirus
N	nucleocapsid protein
M	membrane protein
E	envelope protein
S	spike glycoprotein
ACE2	angiotensin-converting enzyme-2
DAMPs	damage-associated molecular patterns
RdRp	RNA-dependent RNA polymerase
Mpro	main protease
CADD	computer-aided drug design
QM	quantum mechanics
DFT	density functional theory
QSAR	quantitative structure-activity relationship
SAR	structure-activity relationship
SBDD	structure-based drug design
LBDD	ligand-based drug design
AI	artificial intelligence
DL	Deep Learning
GPUs	Graphics Processing Units
CNNs	Convolutional Neural Networks
RNNs	Recurrent Neural Networks
LSTM	Long Short-Term Memory
MTL	Multi-Task Learning