Accelerating drug discovery for Disease X via an AlphaFold2 driven drug repositioning strategy

Abstract

With global concerns about possible outbreaks of Disease X similar to COVID-19 in the future, it is crucial to develop and refine effective drug discovery strategies. Here, we reported a drug discovery strategy for Disease X combined with the FDA-approved drug database, AlphaFold (AF) predicted structures, and virtual screening of integrated our developed molecular docking method (MDCC). A case study of revisiting the main protease (M^pro) of SARS-CoV-2 demonstrated that the lead drug can be discovered through this strategy. First, the M^pro–peptidyl substrate complex structure was predicted using AF2 based on the amino acid sequences of M^pro and its peptidyl substrate. Then, the AF2 predicted structure was optimized through molecular dynamics simulations employing a peptidyl substrate-induced binding pocket approach. Binding free energy analysis revealed that the optimized M^pro exhibited enhanced affinity for the peptidyl substrate, with a ΔG_bind of −103.8 kcal mol⁻¹. Subsequently, this optimized complex structure was used to perform MDCC-based virtual screening of 2005 FDA-approved drugs, from which six drug candidates were selected for M^pro activity determination. Ultimately, Goserelin was identified as a lead compound against M^pro, demonstrating a 75% inhibition rate. Further IC₅₀ determination yielded values of 3.79 μM (pH = 7.5) and 2.05 μM (pH = 6.6), comparable to several reported noncovalent M^pro inhibitors (X77, MCULE-5948770040, and ML188). This work provided a feasible drug discovery strategy in response to the Disease X.

---

Introduction

Disease X defined by the World Health Organization represents the knowledge that a serious international epidemic could be caused by a pathogen currently unknown to cause human diseases.¹ In the 21^st century, several infectious diseases have caused significant damage to global health, economy, and society, such as the 2003 SARS-CoV outbreak, 2009 swine flu pandemic, 2012 MERS-CoV outbreak, 2013–2016 West Africa Ebola epidemic, 2015 Zika virus epidemic, and COVID-19 pandemic, each have been the Disease X of their time.^2,3 In order to effectively respond to these threats, countries have established specialized organizational structures to coordinate and strengthen prevention and control efforts, such as the U.S. Centers for Disease Control and Prevention,⁴ the European Center for Disease Prevention and Control,⁵ the Chinese Center for Disease Control and Prevention,⁶ and the Singapore Ministry of Health.⁷ COVID-19 will not be the last Disease X, and the next Disease X could appear at any time,³ which reminds us to actively prepare for epidemic prevention and quickly master successful drug discovery and development strategies.

Among the pathogens that cause these infectious diseases, coronaviruses occupy a central position in the current global health crisis; especially, SARS-CoV-2, which caused COVID-19, poses a significant challenge to public health.⁸ Coronaviruses belong to the Coronaviridae family, which are enveloped positive-sense RNA viruses with the largest single-stranded RNA genome among all RNA viruses.^9,10 An important step in the lifecycle of coronaviruses is to utilize the viral proteases encoded by the coronavirus genome, the main protease (M^pro, also referred to as 3-chymotrypsin-like protease or 3CL^pro) and papain-like protease (PL^pro), to recognize and cleave the amino acid sequences of virally expressed polyproteins, thereby hydrolyzing the polyproteins to release functional proteins required for genome replication and transcription.^11,12 Therefore, viral proteases are crucial for the processing and maturation of coronaviruses, making them attractive targets for antiviral drug design.

To accelerate the drug discovery process, drug repositioning is now becoming an increasingly important strategy for researchers compared to traditional drug discovery strategies due to its advantages of reducing development costs and risks, shortening time to the market, and improving success rates.^13–15 With the development of computational drug design, virtual screening is important for drug repositioning as it can quickly characterize and optimize new candidate drugs, thereby accelerating drug design and development.^16,17 To help simplify the search, molecular docking has been widely used in virtual screening, especially when the three-dimensional structures of proteins are available.^18,19 However, most protein structures currently used for molecular docking are crystal structures resolved through experiments^20,21 but determining protein structures typically requires months to years of an arduous effort, which is still too long for us who want to overcome Disease X in a short period of time. It is encouraging that compared to three-dimensional structures, the amino acid sequences of proteins can be obtained within weeks or even days of the spread of Disease X.^22,23 Since the advent of the AlphaFold (AF) series models, the field of rapidly predicting protein structures based solely on amino acid sequences has shown remarkable progress.^24–26 Therefore, combined with drug repositioning, docking-based virtual screening, and AF models might be beneficial for accelerating the drug discovery process.

It is worth noting that AF modelled conformational states within the binding sites associated with ligand binding may still not be correct or comprehensive in some cases.^26,27 This might be due to the fact that the binding sites change with the ligand structures,²⁸ but AF models typically predict static structures rather than the dynamic behavior of biomolecular systems.²⁶ However, even small side chain variations in the protein structures used may cause deviations in ligand binding conformation, which can have a significant impact on the docking results for molecular docking.²⁷ Therefore, in order to facilitate the utilization of AF models in docking-based virtual screening, it seems important to enhance the accuracy of the AF model in predicting binding site conformations and molecular docking in predicting ligand conformations. For the former, several studies suggested that ligand-steered co-modeling and co-refinement of protein structures together with their binding ligands might be the preferred strategy for obtaining reasonable binding site conformations if ligands are available.^27,29 While, for the latter, our group developed a docking method, molecular docking with a conformer-dependent charge (MDCC), during the long-term development of theoretical methods for drug and material design.^30–33 This MDCC method has demonstrated relatively high docking accuracy in predicting protein–ligand complexes³² and has been successfully applied to the discovery of SARS-CoV-2 RdRp/JAK1 dual target drugs.³⁴

Here, we reported a drug discovery strategy for Disease X, which combined with the FDA-approved drug database, AF predicted structures, and virtual screening of the integrated MDCC method (Fig. 1), and SARS-CoV-2 M^pro was selected as the revisiting case to demonstrate the application of this strategy. Research has shown that the M^pro inhibitor nirmatrelvir could bind to M^pro by mimicking the peptidyl substrate. However, the binding affinity of nirmatrelvir decreased due to the partial loss of hydrogen bonds and steric hindrance after M^pro underwent mutation. In contrast, the impact of mutations on peptidyl substrate binding was relatively minor.³⁵ Therefore, in this study, we adopted a peptidyl substrate-induced binding pocket approach to obtain the M^pro–peptidyl substrate complex structure using the known amino acid sequences of M^pro and its peptidyl substrate. Subsequently, this structure was utilized for MDCC-based virtual screening to identify drugs with structural and binding modes similar to the peptidyl substrate. Through two rounds of screening and fluorescence resonance energy transfer (FRET) assay analysis, a lead drug, Goserelin, with inhibitory activity against M^pro was discovered among six drug candidates, with IC₅₀ values of 3.79 μM (pH = 7.5) and 2.05 μM (pH = 6.6). Further MD simulations showed that the binding mode of M^pro in complex with Goserelin was similar to those of several reported noncovalent M^pro inhibitors and the peptidyl substrate of M^pro. All these studies on the M^pro case have demonstrated the feasibility of our reported strategy in screening effective drugs against the Disease X, as well as the accuracy in predicting protein–ligand complexes, providing a promising direction for drug discovery in response to the Disease X.

Fig. 1 The drug discovery strategy for Disease X.

Methods

Protein prediction, preparation and refinement

According to the report in the literature, the amino acid sequences of M^pro and its peptidyl substrate can be obtained approximately one week after the discovery of SARS-CoV-2.²³ The structure of SARS-CoV-2 M^pro in complex with the peptidyl substrate was predicted by AF2 based on the amino acid sequences. Then, the Protein Preparation Wizard module of the Schrödinger software³⁶ was used to add hydrogen atoms, assign partial charges and protonation states, and minimize the structure with the OPLS_2005 force field³⁷ to prepare the complex structure predicted by AF2. Subsequently, 200 ns MD simulations were used to refine the complex structure (the specific details of MD simulations were elaborated in the next section). Finally, the Receptor Grid Generation program of the Schrödinger software was used to define the docking grid coordinate for the M^pro structure by centering on the substrate. The grid box was centered at coordinates (6.9, 18.4, −4.4), with an inner box size of 10 × 10 × 10 Å and an outer box size of 42 × 42 × 42 Å.

MD simulations

Molecular dynamics (MD) simulations were carried out using AMBER 22 software.³⁸ The GAFF³⁹ and ff14SB force field⁴⁰ were used for treating ligands and proteins, respectively. The partial atomic charges for ligands were calculated by B3LYP/6-311G(d,p) basis sets using Gaussian 16 software,⁴¹ and the original BCC charges were replaced by RESP charges obtained using Multiwfn 3.8 software.⁴² Then, the complexes were neutralized by adding counterions and placed in a rectangular periodic box of water molecules with the TIP3P⁴³ water model. The distance between the solute and the box boundary was set at 10 Å. In addition, all simulations were conducted under the NPT ensemble at 300.0 K and 1 atm. The snapshots were saved at 10 ps intervals. Finally, the systems were subjected to 200 ns MD simulations in the NPT ensemble with the PMEMD program. The root mean square deviation (RMSD) value of the heavy atoms and hydrogen bond analysis were determined using the cpptraj module included in AmberTools. MD simulations were performed in triplicate for each system. The MD simulations were conducted on a server configured with 16 core CPUs and 4 NVIDIA 4080 GPUs, and took approximately 22 hours.

Docking-based virtual screening

A library with 2005 FDA-approved drugs from Drugbank⁴⁴ was used for docking-based virtual screening. Three-dimensional ligand structures were generated by the LigPrep module of the Schrödinger software, with the OPLS_2005 force field. Docking-based virtual screening was performed by the Glide program of the Schrödinger software using the docking grid prepared in the previous section. As the first filter, standard precision (SP) docking was performed to filtering out 32 drugs. Then, 80 drugs with docking scores of less than −8 kcal mol⁻¹ from the SP docking were selected for secondary filtration using the MDCC method (the specific details of the MDCC method were elaborated in the next section). Finally, the top 20 drugs in the MDCC results were retained. Both SP docking and MDCC docking were performed on the 64 core server. SP docking took about 30 minutes and MDCC docking took about 6.5 hours.

MDCC method

Firstly, 80 drug structures were considered in the geometry optimization and vibrational frequencies at the B3LYP/6-311G(d,p) level combined with DFT-D3 (BJ) correction using Gaussian 16 software to obtain the lowest energy initial structures. Second, multiple drug conformers were generated using the OPLS_2005 force field by the Conformational Search module of the Schrödinger software based on the initial structure obtained for each drug. The top ten conformers of each drug were selected based on the lowest energy principle. Then, all the conformers were subjected to single point energy calculations using the same functional and basis set as described in the first step. The RESP charges were generated using Multiwfn 3.8 software for all these conformers. The original charge values were replaced with these RESP charges in these drug conformer files. Next, molecular docking simulations were performed using the Glide SP method of the Schrödinger software. The RESP charges of all drug conformers were used for molecular docking. Finally, the generated docking poses and the binding energies were scored and ranked using the GlideScore. The highest GlideScore value (with the highest absolute value) was the final docking score of the drug.

SARS-CoV-2 M^pro activity determination

FRET protease assays were used to measure the inhibitory activity of drugs against SARS-CoV-2 M^pro. The assay measured proteolytic cleavage of a fluoregenic M^pro substrate over time. After substrate addition, kinetics of the reaction was measured at excitation/emission = 325/393 nm using a microplate reader (Synergy2), and continuous data collection was performed. For the inhibition rate determination, a substrate solution with a single concentration of 10 μM was added to each well to initiate the reaction. Compound GC376 was selected as the control group. For half-maximal inhibitory concentration (IC₅₀) determination, a substrate solution with a final concentration of 50 μM (2-fold titration) was added to each well to initiate the reaction. Dose-dependent inhibition and IC₅₀ analyses were performed using GraphPad Prism software. The entire process took about six days.

Results

Refining the binding pocket predicted by AF2 through MD simulations

SARS-CoV-2 M^pro cleaves 11 conserved sites on the polyprotein (peptidyl substrate), and its recognition sequence at these sites is x-(L/F/V)-Q↓-(A/S/N)-x (x represents any amino acid; ↓ represents the cleavage site).^45,46 Here, we chose to combine the SARS-CoV-2 M^pro sequence with the peptidyl substrate sequence SAVLQSGFRKM to predict the complex structure of M^pro–peptidyl substrate via AF2. Subsequently, MD simulations were used to refine the complex structure of M^pro–peptidyl substrate predicted by AF2. MD simulations were performed in triplicate, and it was found that the RMSD plots of protein heavy atoms in all three systems tended to stabilize during 200 ns MD simulations, indicating that all three systems achieved structural stability (Fig. S1). By calculating the binding free energies of three systems (−94.3 ± 5.1 kcal mol⁻¹, −85.7 ± 6.8 kcal mol⁻¹, and −103.8 ± 5.7 kcal mol⁻¹), the representative structure from the system with the lowest binding free energy was extracted by conformational clustering for analysis.

Residue Thr26 was the largest contributor to binding free energy by performing MM/GBSA calculations on M^pro residues. In addition, residues Gln19, Thr190, Glu166, Gln189, Cys145, His164, Gln69, Gly143, His163, Asn119, and Phe140 were also key contributors (Table S1). These residues formed some hydrogen bonds with the substrate throughout the simulation process (Table 1), with some occupying nearly or even over 90% in all frames, for example, Glu166(O)-sub(N3H), Thr26(O)-sub(N8H), Glu166(NH)-sub(O4), His164(O)-sub(N5H), Cys145(NH)-sub(O6), Asn119(ND2H)-sub(O11), Thr26(NH)-sub(O10), Gln19(OE1)-sub(N14H), and His163(NE2H)-sub(O7) (Table 1 and Fig. 2a), indicating their importance in stabilizing the binding pocket conformation. Moreover, residues Gln19, Thr26, Gln69, and Asn119 formed some new hydrogen bonds with the substrate after MD optimization, including Gln69(NE2H)-sub(N13), Asn119(O)-sub(N15H), Gln19(NE2)-sub(N13H), Asn119(ND2H)-sub(N14), Thr26(OG1H)-sub(N10), Gln69(O)-sub(N15H), and Gln69(OE1)-sub(N13H) (Table 1). These hydrogen bonds contributed to the formation of a more stable binding pocket structure, and the binding affinity between M^pro and the substrate was also improved, showing a ΔG_bind of −103.8 kcal mol⁻¹. And the MD optimized structure and AF2 predicted structure were compared with the crystal structure (PDB code: 7DVP); it was found that the binding pocket conformation of the complex was further optimized after MD simulations (Fig. 2b and c). The RMSD value of the substrate decreased by about 0.5 Å and the orientation of some residue structures was closer to the crystal structure, indicating that inducing the protein structure through the substrate can obtain a more stable and reasonable protein–substrate complex structure compared to the AF2 predicted structure. This optimized M^pro structure after MD simulations would also be used for the following docking studies.

Table 1 Key hydrogen bonds of M^pro residue–substrate in the M^pro–substrate complexes predicted by AF2 and MD simulations

M^pro residue–substrate	Hydrogen bond distance (Å) (AF2 structure)	Hydrogen bond distance (Å) (MD structure)	Hydrogen bond occupancy rate in all frames
a The numbers represent the atomic numbers of the atoms in the ligand. b The naming convention of this atom that distinguishes it from other elements of the same type in its residue. c The bold numbers represent the distance that can form hydrogen bonds.
Glu166(O)-sub(N3^aH)	2.92	2.93	0.99
Thr26(O)-sub(N8H)	3.14	2.94	0.99
Glu166(NH)-sub(O4)	2.96	2.98	0.99
His164(O)-sub(N5H)	4.42	3.00	0.98
Cys145(NH)-sub(O6)	3.20	3.35	0.96
Asn119(ND2^bH)-sub(O11)	4.86	2.90	0.96
Thr26(NH)-sub(O10)	3.06	3.09	0.92
Gln19(OE1^b)-sub(N14H)	6.19	3.30	0.89
His163(NE2^bH)-sub(O7)	2.91	2.85	0.82
Thr190(O)-sub(N2H)	2.81	2.79	0.74
Gly143(NH)-sub(O6)	3.05	2.98	0.72
Gly143(NH)-sub(O8)	3.30	3.06	0.69
Phe140(O)-sub(N6H)	3.14	3.40	0.68
Gln189(OE1^b)-sub(N4H)	3.07	2.97	0.60
Gln69(NE2^bH)-sub(N13)	12.82	2.90	0.10
Asn119(O)-sub(N15H)	4.76	3.23	0.08
Gln19(NE2^b)-sub(N13H)	8.93	3.16	0.05
Asn119(ND2^bH)-sub(N14)	6.46	3.21	0.03
Thr26(OG1^bH)-sub(N10)	4.29	3.19	0.02
Gln69(O)-sub(N15H)	8.60	3.19	0.02
Gln69(OE1^b)-sub(N13H)	9.01	3.18	0.02

---

Fig. 2 Structural differences of SARS-CoV-2 M^pro–peptidyl substrate complexes predicted by AF2 and MD simulations. (a) The hydrogen bonds with occupancy rates close to or exceeding 90% in all frames in the AF2 predicted structure and MD optimized structure. The peptidyl substrates of the AF2 predicted structure and MD optimized structure are shown as orange and green sticks, respectively. The corresponding residues are shown as light orange and light green sticks with spheres, respectively. The corresponding hydrogen bonds are shown as orange and green dashed lines marked with distance, respectively. (b) Superimposition of peptidyl substrates and some residues within the binding pocket in the crystal structure (PDB code: 7DVP) and the AF2 predicted structure. The peptidyl substrate and residues of the crystal structure are shown as cyan sticks and cyan sticks with spheres, respectively. (c) Superimposition of peptidyl substrates and some residues within the binding pocket in the crystal structure and the MD optimized structure.

Docking-based virtual screening

The docking-based virtual screening was performed using the M^pro structure optimized through MD simulations (Fig. 3). A library with 2005 FDA-approved drugs from Drugbank was prepared and docked to the substrate binding pocket of the MD optimized M^pro. The initial SP docking reduced the number of drugs from 2005 to 1973. Then, the top 81 drugs with docking scores less than −8 kcal mol⁻¹ from the SP step were planned to be selected for the MDCC method screening. Based on the application rule of the MDCC method we previously proposed,³² when the ligand simultaneously met three conditions (the total hydrophobic surface area is >438, the number of hydrophobic atoms is >10, and the molecular weight is >235), the docking success rate of the MDCC method (92.9%) surpassed that of the Glide SP method (71.6%), indicating that such ligands were more suitable for the MDCC method. It was found that 80 out of the top 81 drugs from the previous SP step simultaneously met the three conditions (Table S2). Consequently, these 80 drugs were subjected to secondary screening using the MDCC method (Table S3). Then, a focused analysis was conducted on the top 20 drugs in the MDCC results (Table 2). It is noteworthy that, among these 20 drugs, Cangrelor (IC₅₀ = 0.9 mM⁴⁷) and Saquinavir (IC₅₀ = 31.4 μM⁴⁸) have been studied for their inhibitory activity against M^pro. To identify more effective drugs, these two drugs that have already undergone activity studies were excluded from the screening results. As mentioned earlier, the natural substrates of M^pro are the polyproteins, and its active site exhibits high specificity for peptide bonds. Therefore, we prioritized peptide-based or amide-containing compounds to mimic the structural features of the peptidyl substrate, including Goserelin, Etelcalcetide, Sincalide, Desmopressin, Lypressin, Pentagastrin, Protirelin, Cefoperazone, Carbetocin, and Piperacillin. After considering the binding mode, availability, and commercial relevance for these drugs, six drug candidates (Goserelin, Lypressin, Pentagastrin, Cefoperazone, Carbetocin, and Piperacillin) were then purchased for further SARS-CoV-2 M^pro activity determination.

Fig. 3 Workflow of ensemble docking-based virtual screening of FDA-approved drugs from Drugbank targeting SARS-CoV-2 M^pro.

Table 2 Information on the top 20 drugs in the MDCC results

Drug ID	Name	Docking score	CAS number
a Drugs that have been studied for M^pro activity.
DB00014	Goserelin	−12.71	65807-02-5
DB12865	Etelcalcetide	−11.98	1262780-97-1
DB09142	Sincalide	−11.82	25126-32-3
DB00035	Desmopressin	−11.41	16679-58-6
DB14642	Lypressin	−11.28	50-57-7
DB00183	Pentagastrin	−11.21	5534-95-2
DB09487	Iotrolan	−10.97	79770-24-4
DB03147	Flavin adenine dinucleotide	−10.76	146-14-5
DB06441^a	Cangrelor	−10.68	163706-06-7
DB03247	Flavin mononucleotide	−10.47	146-17-8
DB01249	Iodixanol	−10.24	92339-11-2
DB00492	Fosinopril	−10.20	98048-97-6
DB09158	Trypan blue	−10.15	2538-83-2
DB01232^a	Saquinavir	−10.11	127779-20-8
DB08934	Sofosbuvir	−10.09	1190307-88-0
DB09421	Protirelin	−10.03	24305-27-9
DB01329	Cefoperazone	−9.75	62893-19-0
DB01282	Carbetocin	−9.67	37025-55-1
DB00319	Piperacillin	−9.59	61477-96-1
DB06636	Isavuconazonium	−9.59	742049-41-8

---

Inhibitory effects of drugs on the SARS-CoV-2 M^pro activity

The inhibitory activities of six drug candidates were measured using FRET-based enzymatic assays. Firstly, the inhibition rates of six drugs were determined at a concentration of 10 μM. It can be observed that, among the six drugs, Goserelin showed a significantly higher inhibition rate against SARS-CoV-2 M^pro, reaching 75%, while the inhibition rates of the other five drugs were less than 50% (Fig. 4a). Then, IC₅₀ determination of Goserelin was further carried out. It could be seen that Goserelin displayed potent inhibition of SARS-CoV-2 M^pro activity, with an IC₅₀ value of 3.79 μM (Fig. 4b), which is equivalent to or even better than the inhibitory activity of several reported noncovalent inhibitors of M^pro, including X77 (IC₅₀ = 2.8 μM),⁴⁹ MCULE-5948770040 (IC₅₀ = 4.2 μM),⁵⁰ and ML188 (IC₅₀ = 6.7 μM).⁴⁹

Fig. 4 Inhibitory activity profiles of Goserelin, Lypressin, Pentagastrin, Cefoperazone, Carbetocin, and Piperacillin against SARS-CoV-2 M^pro. (a) Inhibition rates of Goserelin, Lypressin, Pentagastrin, Cefoperazone, Carbetocin, Piperacillin, and the control group (GC376) at 10 μM. (b) Dose-dependent curve of Goserelin at pH 7.5. (c) Dose-dependent curve of Goserelin at pH 6.6. The assays are run in two replicates.

A previous study has shown that virus replication in the nasal turbine can be inhibited by M^pro inhibitor treatment,⁵¹ indicating that M^pro inhibitors may inhibit M^pro activity in the nasal cavity. Therefore, the IC₅₀ value of Goserelin against M^pro was also determined under the pH condition required for the nasal environment (pH = 6.6⁵²). Surprisingly, Goserelin not only maintained inhibitory activity against M^pro at pH 6.6, but its inhibitory ability was significantly enhanced compared to pH 7.5, with an IC₅₀ value being reduced to 2.05 μM and a smaller deviation (Fig. 4c). This phenomenon also indicated that the efficacy of Goserelin in nasal cavity may be more stable, which made it a potential nasal spray to inhibit virus replication in the nasal cavity and prevent further infection.

The above experimental results not only demonstrated the inhibitory activity of Goserelin against SARS-CoV-2 M^pro, but also provided strong evidence for the ability of our reported strategy to screen effective drugs, which combined the FDA-approved drug database, AF predicted structures, and virtual screening of the integrated MDCC method.

The binding mode of SARS-CoV-2 M^pro in complex with Goserelin

As demonstrated in the previous section, the docking results and M^pro activity assays revealed that Goserelin exerted strong inhibition against SARS-CoV-2 M^pro. Goserelin belongs to the class of organic compounds known as oligopeptides, which are organic compounds containing a sequence of between three and ten alpha-amino acids joined by peptide bonds (Fig. 5a). Subsequently, to understand the dynamic binding behavior of SARS-CoV-2 M^pro in complex with Goserelin, 200 ns explicit solvent MD simulations were carried out on the complex structure obtained by the MDCC method. The RMSD plots of protein heavy atoms revealed that the M^pro structure achieved structural stability during 200 ns MD simulations (Fig. 5b).

Fig. 5 The binding mode of SARS-CoV-2 M^pro in complex with Goserelin. (a) The molecular structure of Goserelin. (b) RMSD of heavy atoms of the M^pro structure during 200 ns MD simulations. (c) Cartoon representation of the structure of M^pro in complex with Goserelin. Goserelin is shown as magenta sticks. M^pro is shown as blue ribbons. (d) Surface representation of the binding pocket of M^pro. Four sites: S1, S2, S4, and S1′ are labeled. (e) Surface representation of the partial structure of Goserelin embedded in the M^pro binding pocket. (f) Schematic diagram of M^pro–Goserelin interactions. Goserelin is shown as magenta lines. The residues involved in hydrogen bond formation are shown as blue lines. Hydrophobic residues are shown as gray eyelash-shaped symbols. Hydrogen bonds are shown as green dashed lines. The residue Cys145 is highlighted by a red box. Four sites are labeled. (f) is created by LigPlot⁺.⁵⁵

The electron density map clearly showed that Goserelin exhibited an extended conformation in the binding pocket of SARS-CoV-2 M^pro (Fig. 5c). It has been reported that the active sites of M^pro are highly conserved among all coronavirus M^pros, typically consisting of four sites: S1, S2, S4, and S1′.⁵³ It can be observed that certain structural groups of Goserelin fit well into these four sites (Fig. 5d and e). At the S1 site, the tyrosine group of Goserelin was well embedded within it. The oxygen atom of the tyrosine group formed a 3.06 Å hydrogen bond with the side chain of residue His163. Moreover, the tert-butyl-serine and leucine moieties of Goserelin deeply inserted into the S2 and S4 sites, respectively. The NH donor of the backbone chain of the leucine group also formed a tight hydrogen bond with the carbonyl oxygen atom of the side chain of residue Gln189. The site at S1′ was shallow and large, allowing the serine group of Goserelin to be well surrounded within it. Additionally, the NH donor and the carbonyl oxygen atom of residue Thr26 participated in stabilizing the structure by forming two hydrogen bonds with the serine group (Fig. 5e and f).

In addition to these deeply buried structural groups at the four sites, other moieties of Goserelin also interacted with some residues in the pocket, contributing to the stability of the structure (Fig. 5f). The arginine group of Goserelin located next to the S4 site was stabilized by residue Glu166 through a 3.32 Å hydrogen bond. The adjacent end capped protective group also formed a hydrogen bond with the side chain of residue Asn142. Besides, the arginine, proline and the end protective groups were surrounded by residues Leu141, Leu167, Pro168, and Gln192, producing extensive hydrophobic interactions. While the tryptophan, histidine and pyroglutamic groups at the other end of the Goserelin structure also formed three hydrogen bonds with residues Asn142 and Asn119, making the entire molecular structure more tightly bound within the pocket. Furthermore, it can be clearly seen that residue Cys145 also made the hydrophobic interaction with Goserelin (Fig. 5f). As reported, the strong binding inhibitors of SARS-CoV-2 M^pro typically form tight hydrogen bonds or nonbonded contacts with residue Cys145.⁵⁴

Interestingly, the binding mode of Goserelin in the SARS-CoV-2 M^pro complex structure was similar to those of several reported noncovalent inhibitors (X77, MCULE-5948770040, and ML188), and the peptidyl substrate mentioned earlier (Fig. 6a). Some researchers have pointed out that these three pharmacophore elements played critical roles in molecular binding, that is, the lipophilic site in the S2 pocket, the acceptor site with the side chain NH donor of His163 in the S1 pocket, and the acceptor site with the Glu166 backbone chain NH.⁵⁶ It can be seen that the tert-butyl-serine moieties of Goserelin, the tert-butylanilido group of X77, the dichlorobenzene group of MCULE-5948770040, the tert-butylanilido group of ML188, and the leucine Group of the peptidyl substrate occupied the hydrophobic S2 site and all of them were lipophilic groups (Fig. 6b). In the S1 site, the tyrosine group of Goserelin, the uracil group of MCULE-5948770040, and the glutamine group of the peptidyl substrate were deeply inserted into it, forming hydrogen bonds with the side chain NH of His163, respectively (Fig. 6c, d, f, and h). Although X77 and ML188 did not form hydrogen bonds with His163, their 3-pyridyl groups also occupied this site (Fig. 6c, e, and g). In addition, all five compounds also formed hydrogen bonds with Glu166 (Fig. 6d–h). Notably, X77, ML188, and the peptidyl substrate directly established hydrogen bonds with the backbone chain NH of Glu166 (Fig. 6e, g, and h), while Goserelin and MCULE-5948770040 interacted via hydrogen bonding with the carbonyl oxygen atom on the backbone and side chain of Glu166, respectively (Fig. 6d and f).

Fig. 6 Comparison of the binding modes of SARS-CoV-2 M^pro in complex with Goserelin, X77, MCULE-5948770040, ML188, and the peptidyl substrate. (a) Surface representation of structures of five compounds embedded in the M^pro binding pocket. Goserelin, X77, MCULE-5948770040, ML188, and the peptidyl substrate are shown as magenta, green, orange, yellow, and cyan sticks, respectively. The S1 and S2 sites are highlighted by a green dashed box and an orange dashed box, respectively. (b) Close-up views of the surface representation of five compounds embedded in the S2 site. (c) Close-up views of the surface representation of five compounds embedded in the S1 site. (d) The close-up view of residues His163 and Glu166 in the Goserelin complex. (e) The close-up view of residues His163 and Glu166 in the X77 complex (PDB code: 6W63). (f) The close-up view of residues His163 and Glu166 in the MCULE-5948770040 complex (PDB code: 7LTJ). (g) The close-up view of residues His163 and Glu166 in the ML188 complex (PDB code: 7L0D). (h) The close-up view of residues His163 and Glu166 in the peptidyl substrate complex (PDB code: 7DVP). The residues His163 and Glu166 in (d)–(h) are shown as blue, wheat, greencyan, purple, and pink sticks with spheres, respectively. Hydrogen bonds are shown as gray dashed lines. S1 sites are labeled.

In summary, the similar binding modes with the peptidyl substrate and three reported noncovalent inhibitors demonstrated the rationality of our given binding mode of SARS-CoV-2 M^pro in complex with Goserelin again, and to some extent provided a molecular level explanation for the strong inhibition of Goserelin against SARS-CoV-2 M^pro described earlier.

Discussion

The reason for choosing M^pro as a revisiting case study is two-fold: firstly, viral proteases are crucial for the processing and maturation of coronaviruses. The viral proteases encoded by the coronavirus genome can cleave the amino acid sequences of virally expressed polyproteins. In addition to coronaviruses, several other viruses also rely on viral proteases to cleave polyproteins, such as the NS2B-NS3 protease of flaviviruses⁵⁷ and the 3C protease of Picornaviridae.⁵⁸ This conserved mechanistic similarity enables the drug design strategies targeting M^pro to be rapidly adapted and applied to combat future Disease X caused by viruses dependent on viral proteases. Another reason is that although a large number of high-quality M^pro structures have been recorded in the PDB bank and AF2 has reached the frontier of the set of experimentally determined structures, it has been found that the AF2 model deviated from experimental structural centroids.⁵⁹ The structural similarity between two randomly selected experimental structures was typically higher than that between the experimental structure and the predicted model. Specifically, the atomic RMSD of the experimental-to-experimental structure was generally smaller than that of experimental-to-AF2, indicating that the various structural states of M^pro and the structural variability of the ligand binding site pose challenges to the accurate prediction of AF2. To improve the accuracy of AF2 in predicting the ligand binding site conformation, the peptidyl substrate was used to induce the ligand binding site through MD simulations, and finally obtained a M^pro–substrate complex structure with a more stable and reasonable binding site conformation compared to the AF2's predicted structure.

Interestingly, Goserelin identified in this work, a gonadotropin releasing hormone receptor agonist used to treat prostate cancer and breast cancer, has also been suggested for repurposing for the treatment of dengue fever⁶⁰ and Candida albicans infection,⁶¹ once again demonstrating the potential of Goserelin in drug repositioning. However, it is worth mentioning that Goserelin may affect women's menstrual cycle, although a study has shown that this effect is a reversible process, and most patients' menstrual cycles gradually return to normal after discontinuing the medication.⁶² Although this represents a potential side effect, Goserelin's reproductive effects are primarily mediated through systemic exposure following subcutaneous administration. In contrast, its potential as a nasal spray may enable low-dose intranasal delivery to directly target the upper respiratory tract, thereby minimizing systemic absorption and improving safety. Related in vivo studies would also be valuable to further validate the feasibility of this hypothesis in future work.

Conclusions

The rapid and effective discovery of drugs for the Disease X remains the top problem that urgently needs to be solved. In this study, we reported a strategy for the rapid discovery of antiviral candidate drugs using an integrated MDCC method that combined the FDA-approved drug database, AF2 predicted structures, MD simulations and virtual screening. A lead drug, Goserelin, screened from 2005 FDA-approved drugs was found to effectively inhibit M^pro (IC₅₀ = 3.79 μM) through SP docking, MDCC docking, and FRET protease assays. Further MD simulations demonstrated that Goserelin was in close contact with the binding pocket of M^pro, and its binding mode was similar to those of three reported noncovalent inhibitors (X77, MCULE-5948770040, and ML188), and the peptidyl substrate of M^pro. This similar binding mode indicated the rationality of the binding mode of Goserelin we proposed, demonstrating the accuracy of integrating AF2 and the MDCC method in predicting protein–ligand complexes.

In summary, the successful discovery of the lead drug for SARS-CoV-2 M^pro has deepened the recognition of our reported strategy, making it possible to become a promising direction for the drug discovery of the Disease X.

Author contributions

W. D. and X. H. designed the study. H. Z., X. H., and W. D. collected, analysed and interpreted the data. W. Q., K. L., and J. Z. assisted in conducting experiments. H. Z. drafted the manuscript. W. D. and X. H. participated in the manuscript revision. All the authors reviewed the manuscript and approved the submitted version.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data supporting this article have been included as part of the SI. See DOI: https://doi.org/10.1039/d5cp01365h

Acknowledgements

This research was supported by the National Natural Science Foundation of China (No. 22273049), the National Key R&D Program of China (No. 2022YFA1503102), the Taishan Scholars Project (No. tspd20230601) and the Development Plan for Youth Innovation Team of Shandong Province (No. 2023KJ011).

References

1. Prioritizing diseases for research and development in emergency contexts. https://www.who.int/activities/prioritizing-diseases-for-research-and-development-in-emergency-contexts, Accessed September 9, 2024.
2. R. E. Baker, A. S. Mahmud, I. F. Miller, M. Rajeev, F. Rasambainarivo, B. L. Rice, S. Takahashi, A. J. Tatem, C. E. Wagner, L.-F. Wang, A. Wesolowski and C. J. E. Metcalf, Nat. Rev. Microbiol., 2021, 20, 193–205.
3. M. D. V. Kerkhove, M. J. Ryan and T. A. Ghebreyesus, Science, 2021, 374, 377.
4. https://www.cdc.gov/index.html, Accessed January 10, 2025.
5. https://www.ecdc.europa.eu/en, Accessed January 10, 2025.
6. https://www.chinacdc.cn/, Accessed January 10, 2025.
7. https://www.moh.gov.sg/, Accessed January 10, 2025.
8. N. Zhu, D. Y. Zhang, W. L. Wang, X. W. Li, B. Yang, J. Song, X. Zhao, B. Huang, W. Shi, R. Lu, P. Niu, F. Zhan, X. Ma, D. Wang, W. Xu, G. Wu, G. F. Gao and W. Tan, New Engl. J. Med., 2020, 382, 727–733.
9. M. M. C. Lai and D. Cavanagh, Adv. Virus Res., 1997, 48, 1–100.
10. P. S. Masters, Adv. Virus Res., 2006, 66, 193–292.
11. V. Anirudhan, H. Lee, H. Cheng, L. Cooper and L. Rong, J. Med. Virol., 2021, 93, 2722–2734.
12. J. Zephyr, N. K. Yilmaz and C. A. Schiffer, Enzymes, 2021, 50, 301–333.
13. S. Naylor, D. M. Kauppi and J. M. Schonfeld, Drug Discovery World, 2015, 16, 57–72.
14. M. Simsek, B. Meijer, A. A. van Bodegraven, N. K. H. de Boer and C. J. J. Mulder, Drug Discovery Today, 2018, 23, 17–21.
15. M. S. Boguski, K. D. Mandl and V. P. Sukhatme, Science, 2009, 324, 1394–1395.
16. E. S. Ozdemir, F. Halakou, R. Nussinov, A. Gursoy and O. Keskin, Methods Mol. Biol., 2019, 1903, 1–21.
17. N. T. Issa, J. Kruger, S. W. Byers and S. Dakshanamurthy, Expert Rev. Clin. Pharmacol., 2014, 6, 95–97.
18. D. A. Gschwend, A. C. Good and I. D. Kuntz, J. Mol. Recognit., 1996, 9, 175–186.
19. X. Y. Meng, H. X. Zhang, M. Mezei and M. Cui, Curr. Comput.-Aided Drug Des., 2011, 7, 146–157.
20. X. J. Jin, M. Zhang, B. B. Fu, M. Li, J. Yang, Z. Zhang, C. Li, H. Zhang, H. Wu, W. Xue and Y. Liu, J. Chem. Inf. Model., 2024, 64, 1319–1330.
21. S. Sharma and S. Deep, J. Biomol. Struct. Dyn., 2020, 40, 3003–3010.
22. S. Baize, D. Pannetier, L. Oestereich, T. Rieger, L. Koivogui, N. Magassouba, B. Soropogui, M. S. Sow, S. Keita, H. De Clerck, A. Tiffany, G. Dominguez, M. Loua, A. Traore, M. Kolie, E. R. Malano, E. Heleze, A. Bocquin, S. Mely, H. Raoul, V. Caro, D. Cadar, M. Gabriel, M. Pahlmann, D. Tappe, J. Schmidt-Chanasit, B. Impouma, A. K. Diallo, P. Formenty, M. Van Herp and S. Gunther, New Engl. J. Med., 2014, 371, 1418–1425.
23. F. Wu, S. Zhao, B. Yu, Y. M. Chen, W. Wang, Z. G. Song, Y. Hu, Z. W. Tao, J. H. Tian, Y. Y. Pei, M. L. Yuan, Y. L. Zhang, F. H. Dai, Y. Liu, Q. M. Wang, J. J. Zheng, L. Xu, E. C. Holmes and Y. Z. Zhang, Nature, 2020, 579, 265–269.
24. J. Jumper, R. Evans, A. Pritzel, T. Green, M. Figurnov, O. Ronneberger, K. Tunyasuvunakool, R. Bates, A. Zidek, A. Potapenko, A. Bridgland, C. Meyer, S. A. A. Kohl, A. J. Ballard, A. Cowie, B. Romera-Paredes, S. Nikolov, R. Jain, J. Adler, T. Back, S. Petersen, D. Reiman, E. Clancy, M. Zielinski, M. Steinegger, M. Pacholska, T. Berghammer, S. Bodenstein, D. Silver, O. Vinyals, A. W. Senior, K. Kavukcuoglu, P. Kohli and D. Hassabis, Nature, 2021, 596, 583–589.
25. R. Evans, M. O’Neill, A. Pritzel, N. Antropova, A. Senior, T. Green, A. Žídek, R. Bates, S. Blackwell, J. Yim, O. Ronneberger, S. Bodenstein, M. Zielinski, A. Bridgland, A. Potapenko, A. Cowie, K. Tunyasuvunakool, R. Jain, E. Clancy, P. Kohli, J. Jumper and D. Hassabis, bioRxiv, preprint, 2021.10.04.463034, 2022 DOI:10.1101/2021.10.04.463034.
26. J. Abramson, J. Adler, J. Dunger, R. Evans, T. Green, A. Pritzel, O. Ronneberger, L. Willmore, A. J. Ballard, J. Bambrick, S. W. Bodenstein, D. A. Evans, C.-C. Hung, M. O’Neill, D. Reiman, K. Tunyasuvunakool, Z. Wu, A. Žemgulytė, E. Arvaniti, C. Beattie, O. Bertolli, A. Bridgland, A. Cherepanov, M. Congreve, A. I. Cowen-Rivers, A. Cowie, M. Figurnov, F. B. Fuchs, H. Gladman, R. Jain, Y. A. Khan, C. M. R. Low, K. Perlin, A. Potapenko, P. Savy, S. Singh, A. Stecula, A. Thillaisundaram, C. Tong, S. Yakneen, E. D. Zhong, M. Zielinski, A. Žídek, V. Bapst, P. Kohli, M. Jaderberg, D. Hassabis and J. M. Jumper, Nature, 2024, 630, 493–500.
27. V. Scardino, J. I. Di Filippo and C. N. Cavasotto, iScience, 2023, 26, 105920.
28. P. Csermely, R. Palotai and R. Nussinov, Trends Biochem. Sci., 2010, 35, 539–546.
29. S. S. Phatak, E. A. Gatica and C. N. Cavasotto, J. Chem. Inf. Model., 2010, 50, 2119–2128.
30. L. Yang, L. Sun, Y. Zhao, J. Sun, Q. Deng, H. Wang and W. Deng, Phys. Chem. Chem. Phys., 2021, 23, 22835–22853.
31. X. Hu, J. Pang, J. Zhang, C. Shen, X. Chai, E. Wang, H. Chen, X. Wang, M. Duan, W. Fu, L. Xu, Y. Kang, D. Li, H. Xia and T. Hou, Adv. Sci., 2021, 9, 2102435.
32. H. Zhao, L. Sun, D. Zhang, X. Hu and W. Deng, Phys. Chem. Chem. Phys., 2024, 26, 22598–22610.
33. L. Yang, L. Sun and W.-Q. Deng, J. Phys. Chem. A, 2018, 122, 1672–1677.
34. H. Zhao, X. Chang, L. Sun, E. Li, D. Zhang, W. Qi, J. Chai, X. Hu and W. Deng, J. Phys. Chem. B, 2024, 128, 12066–12077.
35. Y. Zhou, K. A. Gammeltoft, L. A. Ryberg, L. V. Pham, H. D. Tjørnelund, A. Binderup, C. R. D. Hernandez, C. Fernandez-Antunez, A. Offersgaard, U. Fahnøe, G. H. J. Peters, S. Ramirez, J. Bukh and J. M. Gottwein, Sci. Adv., 2022, 8, eadd7197.
36. L. Schrödinger, Release 2018-4, Schrödinger, New York, NY, 2018.
37. G. A. Kaminski, R. A. Friesner, J. Tirado-Rives and W. L. Jorgensen, J. Phys. Chem. B, 2001, 105, 6474–6487.
38. D. A. Case, H. M. Aktulga, K. Belfon, I. Y. Ben-Shalom, J. T. Berryman, S. R. Brozell, D. S. Cerutti, T. E. Cheatham, III, G. A. Cisneros, V. W. D. Cruzeiro, T. A. Darden, R. E. Duke, G. Giambasu, M. K. Gilson, H. Gohlke, A. W. Goetz, R. Harris, S. Izadi, S. A. Izmailov, K. Kasavajhala, M. C. Kaymak, E. King, A. Kovalenko, T. Kurtzman, T. S. Lee, S. LeGrand, P. Li, C. Lin, J. Liu, T. Luchko, R. Luo, M. Machado, V. Man, M. Manathunga, K. M. Merz, Y. Miao, O. Mikhailovskii, G. Monard, H. Nguyen, K. A. O’Hearn, A. Onufriev, F. Pan, S. Pantano, R. Qi, A. Rahnamoun, D. R. Roe, A. Roitberg, C. Sagui, S. Schott-Verdugo, A. Shajan, J. Shen, C. L. Simmerling, N. R. Skrynnikov, J. Smith, J. Swails, R. C. Walker, J. Wang, J. Wang, H. Wei, R. M. Wolf, X. Wu, Y. Xiong, Y. Xue, D. M. York, S. Zhao and P. A. Kollman, Amber 2022, University of California, San Francisco, 2022.
39. J. M. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman and D. A. Case, J. Comput. Chem., 2004, 25, 1157–1174.
40. J. A. Maier, C. Martinez, K. Kasavajhala, L. Wickstrom, K. E. Hauser and C. Simmerling, J. Chem. Theory Comput., 2015, 11, 3696–3713.
41. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone, G. A. Petersson, H. Nakatsuji, X. Li, M. Caricato, A. V. Marenich, J. Bloino, B. G. Janesko, R. Gomperts, B. Mennucci, H. P. Hratchian, J. V. Ortiz, A. F. Izmaylov, J. L. Sonnenberg, D. Williams-Young, F. Ding, F. Lipparini, F. Egidi, J. Goings, B. Peng, A. Petrone, T. Henderson, D. Ranasinghe, V. G. Zakrzewski, J. Gao, N. Rega, G. Zheng, W. Liang, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. Vreven, K. Throssell, J. A. Montgomery, Jr., J. E. Peralta, F. Ogliaro, M. J. Bearpark, J. J. Heyd, E. N. Brothers, K. N. Kudin, V. N. Staroverov, T. A. Keith, R. Kobayashi, J. Normand, K. Raghavachari, A. P. Rendell, J. C. Burant, S. S. Iyengar, J. Tomasi, M. Cossi, J. M. Millam, M. Klene, C. Adamo, R. Cammi, J. W. Ochterski, R. L. Martin, K. Morokuma, O. Farkas, J. B. Foresman and D. J. Fox, Gaussian 16, Gaussian, Inc., Wallingford CT, 2016.
42. T. Lu and F. W. Chen, J. Comput. Chem., 2012, 33, 580–592.
43. V. Hornak, R. Abel, A. Okur, B. Strockbine, A. Roitberg and C. Simmerling, Proteins: Struct., Funct., Bioinf., 2006, 65, 712–725.
44. D. S. Wishart, C. Knox, A. C. Guo, S. Shrivastava, M. Hassanali, P. Stothard, Z. Chang and J. Woolsey, Nucleic Acids Res., 2006, 34, 668–672.
45. E. Cho, M. Rosa, R. Anjum, S. Mehmood, M. Soban, M. Mujtaba, K. Bux, S. T. Moin, M. Tanweer, S. Dantu, A. Pandini, J. Yin, H. Ma, A. Ramanathan, B. Islam, A. S. J. S. Mey, D. Bhowmik and S. Haider, J. Chem. Inf. Model., 2021, 61, 3058–3073.
46. J. Lee, C. Kenward, L. J. Worrall, M. Vuckovic, F. Gentile, A.-T. Ton, M. Ng, A. Cherkasov, N. C. J. Strynadka and M. Paetzel, Nat. Commun., 2022, 13, 5196.
47. A. Gupta, C. Rani, P. Pant, V. Vijayan, N. Vikram, P. Kaur, T. P. Singh, S. Sharma and P. Sharma, ACS Omega, 2020, 5, 33151–33161.
48. M. Mahdi, J. A. Mótyán, Z. I. Szojka, M. Golda, M. Miczi and J. Tőzsér, Virol. J., 2020, 17, 190.
49. A. Luttens, H. Gullberg, E. Abdurakhmanov, D. D. Vo, D. Akaberi, V. O. Talibov, N. Nekhotiaeva, L. Vangeel, S. De Jonghe, D. Jochmans, J. Krambrich, A. Tas, B. Lundgren, Y. Gravenfors, A. J. Craig, Y. Atilaw, A. Sandström, L. W. K. Moodie, Å. Lundkvist, M. J. van Hemert, J. Neyts, J. Lennerstrand, J. Kihlberg, K. Sandberg, U. H. Danielson and J. Carlsson, J. Am. Chem. Soc., 2022, 144, 2905–2920.
50. A. Clyde, S. Galanie, D. W. Kneller, H. Ma, Y. Babuji, B. Blaiszik, A. Brace, T. Brettin, K. Chard, R. Chard, L. Coates, I. Foster, D. Hauner, V. Kertesz, N. Kumar, H. Lee, Z. Li, A. Merzky, J. G. Schmidt, L. Tan, M. Titov, A. Trifan, M. Turilli, H. Van Dam, S. C. Chennubhotla, S. Jha, A. Kovalevsky, A. Ramanathan, M. S. Head and R. Stevens, J. Chem. Inf. Model., 2021, 62, 116–128.
51. B. X. Quan, H. P. Shuai, A. J. Xia, Y. Hou, R. Zeng, X.-L. Liu, G.-F. Lin, J.-X. Qiao, W.-P. Li, F.-L. Wang, K. Wang, R.-J. Zhou, T. T.-T. Yuen, M.-X. Chen, C. Yoon, M. Wu, S.-Y. Zhang, C. Huang, Y.-F. Wang, W. Yang, C. Tian, W.-M. Li, Y.-Q. Wei, K.-Y. Yuen, J. F.-W. Chan, J. Lei, H. Chu and S. Yang, Nat. Microbiol., 2022, 7, 716–725.
52. A. J. B. Kreutzberger, A. Sanyal, A. Saminathan, L. M. Bloyet, S. Stumpf, Z. Liu, R. Ojha, M. T. Patjas, A. Geneid, G. Scanavachi, C. A. Doyle, E. Somerville, R. B. D. C. Correia, G. D. Caprio, S. Toppila-Salmi, A. Mäkitie, V. Kiessling, O. Vapalahti, S. P. J. Whelan, G. Balistreri and T. Kirchhausen, Proc. Natl. Acad. Sci. U. S. A., 2022, 119, e2209514119.
53. H. T. Yang, W. Q. Xie, X. Y. Xue, K. Yang, J. Ma, W. Liang, Q. Zhao, Z. Zhou, D. Pei, J. Ziebuhr, R. Hilgenfeld, K. Y. Yuen, L. Wong, G. Gao, S. Chen, Z. Chen, D. Ma, M. Bartlam and Z. Rao, PLoS Biol., 2005, 3, e324.
54. W. H. Dai, B. Zhang, X. M. Jiang, H. X. Su, J. Li, Y. Zhao, X. Xie, Z. JIn, J. Peng, F. Liu, C. Li, Y. Li, F. Bai, H. Wang, X. Cheng, X. Cen, S. Hu, X. Yang, J. Wang, X. Liu, G. Xiao, H. Jiang, Z. Rao, L.-K. Zhang, Y. Xu, H. Yang and H. Liu, Science, 2020, 368, 1331–1335.
55. R. A. Laskowski and M. B. Swindells, J. Chem. Inf. Model., 2011, 51, 2778–2786.
56. Y. Unoh, S. Uehara, K. Nakahara, H. Nobori, Y. Yamatsu, S. Yamamoto, Y. Maruyama, Y. Taoda, K. Kasamatsu, T. Suto, K. Kouki, A. Nakahashi, S. Kawashima, T. Sanaki, S. Toba, K. Uemura, T. Mizutare, S. Ando, M. Sasaki, Y. Orba, H. Sawa, A. Sato, T. Sato, T. Kato and Y. Tachibana, J. Med. Chem., 2022, 65, 6499–6512.
57. D. Luo, S. G. Vasudevan and J. Lescar, Antivir. Res., 2015, 118, 148–158.
58. R. O. Diarimalala, M. Hu, Y. Wei and K. Hu, Virol. J., 2020, 17, 173.
59. T. J. Lane, Nat. Methods, 2023, 20, 170–173.
60. S. Gautam, A. Thakur, A. Rajput and M. Kumar, Viruses, 2023, 16, 45.
61. A. Gao, V. L. Kouznetsova and I. F. Tsigelny, Mycoses, 2022, 65, 794–805.
62. A. D. Anastasilakis, S. Papachatzopoulos, P. Makras, A. Gkiomisi, P. Nikolakopoulos, S. A. Polyzos, C. Ntenti, I. Ballaouri, S. Gerou, O. Tsachouridou, A. Papatheodorou, K. Aliazis, S. Fermanoglou, I. Bisbinas and M. P. Yavropoulou, Bone, 2022, 158, 116354.

---