Sage Journals: Discover world-class research

Abstract

Objectives: In our continuous efforts to combat COVID-19, our objective was to conduct a comprehensive computer-aided drug design study utilizing 4924 African natural metabolites sourced from diverse databases across various African regions from 1962 to 2019. The primary goal was to target the SARS-CoV-2 helicase, a crucial enzyme in viral replication. Methods: We employed structural fingerprint and molecular similarity studies with VXG, the co-crystallized ligand, as a reference. Subsequently, docking and absorption, distribution, metabolism, excretion, and toxicity (ADMET) studies were conducted. Results: The structural fingerprint analysis identified 200 structurally similar compounds, then the molecular similarity studies selected 40 compounds. Among them, 15 metabolites with low free energies, favorable binding modes, and promising ADMET properties were identified. Four compounds were excluded according to the toxicity studies. Compound 1552, 1-((S)-2,3-dihydro-2-((R)-1-hydroxypropan-2-yl)benzofuran-5-yl)ethenone, exhibited the most favorable docking results. Molecular dynamics simulations conclusively demonstrated its stable binding to the SARS-CoV-2 helicase, characterized by low energy and optimal dynamics. Conclusion: The findings suggest promising avenues for potential COVID-19 cures, encouraging further exploration through in vitro and in vivo studies of the identified compounds, particularly compound 1552.

Keywords

SARS-CoV-2 helicase natural products structure similarity MD simulations MM-GBSA PCAT

Introduction

The WHO had reported a total of 769 806 130 confirmed cases of COVID-19 globally, with 6 955 497 recorded deaths. Despite the administration of 13 499 767 926 vaccine doses worldwide, the virus persists in affecting populations on a significant scale.¹ These concerning statistics act as a clear reminder of the significant challenge confronting scientists globally. The persistent spread of the virus, highlighted by these figures, emphasizes the immediate and collective action needed from the global scientific community to uncover effective treatments and preventive measures.

In the past 20 years, the adoption of computer-aided drug discovery (CADD) methods has seen a significant upswing in pharmaceutical research and experimentation.^2,3 This widespread use can be attributed to several key factors. First, the ability to determine the precise 3D structures of numerous protein targets within the human body.⁴ Second, the remarkable progress in computer software and hardware capabilities.⁵ Lastly, an enhanced comprehension of the principles of structure–activity relationship (SAR) and quantitative structure–activity relationship.⁶ As a result, computational chemistry techniques have been applied to explore various pharmacokinetic and pharmacodynamic properties, enabling the identification of relationships between chemical structure and activity. These methods include pharmacophore identification,⁷ homology modeling,⁸ docking,⁹ molecular dynamics (MD) simulations,¹⁰ absorption, distribution, metabolism, excretion, and toxicity (ADMET) predictions,¹¹ density functional theory (DFT),¹² structure similarity,¹³ and fingerprints.¹⁴

Since the dawn of history, nature has played an essential role in fulfilling basic human requirements. It has not only supplied remedies for medical treatments but has also provided sustenance and played a part in creating a range of cosmetic products.^15,16

In our efforts to combat COVID-19, we employed a multistage in silico CADD approach to identify natural inhibitors and recommend the most suitable one against specific COVID-19 enzymes. From a group of 310 antiviral natural metabolites, potential inhibitors for nsp10,¹⁶ the main protease,^17,18 and the papain-like protease¹⁹ have been pointed. Additionally, among 3009 food drug administration (FDA)-approved drugs, the most promising inhibitors against the RNA-dependent RNA polymerase²⁰ and the SARS-CoV-2 nsp16-nsp10 2′-o-methyltransferase complex²¹ have been indicated. Also, we conducted an exploration of a collection of 5956 compounds sourced from traditional Chinese medicine with the aim of identifying potential natural inhibitors for the SARS-CoV-2 helicase enzyme.²²

We herein report the structure and ligand-based in silico discovery of potential SARS-CoV-2 helicase inhibitors among 4924 natural African metabolites. The screened compounds were obtained from the African Natural Products Database, which was collected from various natural products databases of several African regions covering from 1962 to 2019. The data set was obtained from local and international African journals besides MSc and PhD theses in African universities.²³ The applied methods were structure similarity, structural fingerprints, molecular docking, ADMET, toxicity, MD simulations (Root Mean Square Deviation [RMSD], Root Mean Square Fluctuation [RMSF], R_g, SASA, and H-bonding) for 200 ns in addition to molecular mechanics-generalized Boltzmann surface area (MM-GBSA), protein–ligand interaction profile (PLIP), and principal component analysis of trajectory (PCAT) studies.

Results and Discussion

Structure Fingerprints Studies

The co-crystallized ligand is a compound with a pronounced affinity for a specific protein, resulting in the formation of a crystalline complex. This complex serves as a valuable source of information revealing the precise structure of the protein through x-ray imaging.²⁴ Identify intricate interactions between the ligand and the protein, exposing essential structural and chemical features defining their potent binding. The co-crystallized ligand's chemical structure serves as a valuable blueprint for crafting new inhibitors effective in binding to the target protein, advancing the search for potential therapeutic agents..²⁵ Analyzing the structural attributes and functional groups of the co-crystallized ligand provides a deeper insight into the key elements driving its strong binding affinity. Utilizing structure fingerprints involves a ligand-based computational approach, calculating chemical and physical descriptors, and comparing them against a specific active molecule.²⁶ The fingerprints study was employed vastly to expect the similarity and/or dissimilarity of two molecules or more.^27,28 The software computes the chemical and physical descriptors of the examined ligand as mathematical symbols and converts them to bit strings. The obtained strings calculate the presence (1) as well as the absence (0) of a specific 2D fragment in both examined and reference compounds.^29,30

We applied Discovery Studio software to explore the degree of similarity of the fingerprints of 4924 natural African metabolites with VXG, the co-crystallized of SARS-CoV-2 helicase (PDB ID: 5RMM). The selection of PDB entry (5RMM) was based on the pivotal role of the helicase in the viral replication, coupled with the simplicity of its co-crystallized ligand (VXG). This choice increases the likelihood of identifying structurally similar compounds through the examined library, aligning with the study's objective of exploring potential inhibitors targeting the SARS-CoV-2 helicase. The fingerprints experiment was adjusted to determine the most similar 200 compounds among the examined compounds (Table S1 in Supplementary data). The fingerprints study described the following descriptors in the fragments as well as atoms of the examined compounds and VXG: H-bond acceptors,³¹ H-bond donors,³² charges,³³ hybridization,³⁴ positive ionizable atoms,³⁵ negative ionizable atoms,³⁶ halogens,³⁷ aromatic groups,³⁸ align with the ALogP.³⁹

Molecular Similarity

The main difference between fingerprints and molecular similarity is that the first calculates the presence or the absence of certain 2D atomic descriptors in fragments, or substructures.⁴⁰ On the other hand, the molecular similarity study computes the molecular descriptors regarding the whole structure of the examined compound. The tested descriptors could be steric, electronic, topological, or physical.⁴¹ We depended on Discovery studio software to examine the molecular similarity of the most similar 200 metabolites against VXG. The tested descriptors in this study (Figure 1 and Table 1) included partition coefficient (ALog p),⁴² molecular weight (MW),⁴³ H-bond donors (HBDs),⁴⁴ H-bond acceptors (HBAs),⁴⁵ rotatable bonds number,⁴⁶ number of rings as well as aromatic rings⁴⁷ and minimum distance⁴⁸ together with the molecular fractional polar surface area (MFPSA).⁴⁹ The similarity study was adjusted to single out the most similar 40 compounds (Table 2 and Figure S1 in the Supplementary data).

Figure 1.

The molecular similarity of the most similar 40 natural African metabolites and VXG. The green balls represent the reference ligand (VXG), the red balls represent the most similar compounds with VXG, and the blue balls represent compounds with low similarity with VXG.

Table 1.

Molecular Descriptors of the Most Similar 49 Natural African Metabolites and VXG.

Comp.	ALog p	MW	HBA	HBD	Rotatable bonds	Rings	Aromatic rings	MFPSA	Minimum distance
VXG	0.71	233.26	3	1	2	2	1	0.237
956	1.69	164.16	3	2	2	1	1	0.338	0.702
1031	1.9	146.14	2	0	0	2	1	0.188	0.74
1060	1.46	122.12	2	1	1	1	1	0.291	0.735
1165	1.93	148.16	2	1	2	1	1	0.235	0.65
1166	1.67	194.18	4	2	3	1	1	0.325	0.665
1213	3.08	230.3	2	1	2	2	1	0.149	0.739
1335	2.16	216.23	3	0	2	3	1	0.186	0.618
1337	1.26	174.15	3	0	1	2	1	0.263	0.477
1461	1.42	180.2	3	1	4	1	1	0.236	0.568
1546	1.81	164.2	2	1	3	1	1	0.202	0.62
1549	2.09	220.26	3	2	4	1	1	0.229	0.67
1550	2.09	220.26	3	2	4	1	1	0.229	0.67
1551	2.35	218.25	3	1	4	1	1	0.222	0.654
1552	1.63	220.26	3	1	3	2	1	0.199	0.335
1553	1.63	220.26	3	1	3	2	1	0.199	0.335
1556	2.21	262.3	4	1	5	2	1	0.201	0.7
1557	2.21	262.3	4	1	5	2	1	0.201	0.7
1628	1.3	287.31	4	1	5	2	1	0.215	0.622
1691	1.06	219.19	5	2	2	2	1	0.356	0.62
1944	2.76	264.32	4	2	2	2	1	0.229	0.675
2570	0.87	166.17	3	2	3	1	1	0.329	0.651
2572	2.26	192.21	3	1	4	1	1	0.221	0.664
2584	1.97	190.24	2	0	1	2	1	0.127	0.669
2682	1.7	227.26	2	1	0	3	1	0.167	0.615
2699	1.15	189.21	2	0	1	2	1	0.144	0.551
2717	1.29	221.21	5	2	1	2	1	0.356	0.654
2810	1.26	275.34	4	1	4	3	1	0.178	0.594
2817	1.74	287.35	4	1	4	3	1	0.169	0.671
2980	2.34	261.32	3	0	3	2	1	0.163	0.628
3045	0.96	244.29	2	1	2	3	1	0.205	0.392
3046	0.72	260.29	3	2	2	3	1	0.275	0.471
3076	1.33	208.23	3	0	1	2	1	0.315	0.476
3084	1.06	240.26	3	1	0	4	1	0.228	0.733
3088	0.71	161.16	2	2	0	2	1	0.321	0.66
3148	2.42	246.26	4	2	3	2	1	0.265	0.615
3155	2.03	246.26	4	1	0	3	1	0.229	0.619
4143	2.02	208.21	4	2	4	1	1	0.301	0.723
4253	1.55	150.17	2	0	2	1	1	0.151	0.725
4466	2.36	202.25	2	0	1	2	1	0.117	0.733
4558	1.44	152.15	3	1	2	1	1	0.285	0.575

Table 2.

Chemical Names, Structures, and Binding Free Energies (Calculated ΔG in kcal/mol) of the Most Similar Natural African Metabolites and Ligand VXG Against SARS-CoV-2 Helicase Enzyme.

Comp.	Chemical name	Amino acids	ΔG
VXG	(3S,4R)-1-acetyl-4-phenylpyrrolidine-3-carboxylic acid	Ser486, Asn516, Asn177, Tyr269, Tyr265	−5.24
956	O-coumaric acid	-	−4.62
1031	Coumarin	-	−4.17
1060	Benzoic acid	-	−3.96
1165	Trans-cinnamic acid	Asn516, Asn1177, Ser486, Pro408	−4.55
1166	Trans-ferulic acid	-	−4.66
1213	Catalponol	-	−4.27
1335	(+)-(S,R)-epoxyartemidin	-	−4.23
1337	Artemidinal	-	−4.63
1461	Ethyl mandelate	-	−4.27
1546	4-(p-hydroxyphenyl)-butan-2-one	-	−4.59
1549	1-(4-hydroxy-3-((Z)-4-hydroxy-3-methylbut-2-enyl)phenyl)ethanone	-	−5.08
1550	1-(4-hydroxy-3-((E)-4-hydroxy-3-methylbut-2-enyl)phenyl)ethanone	-	−4.7
1551	(2E)-4-(5-acetyl-2-hydroxyphenyl)-2-methylbut-2-enal	Asn534, Asn177, Asn534, Asn177, Asp534, His554	−4.48
1552	1-((S)-2,3-dihydro-2-((R)-1-hydroxypropan-2-yl)benzofuran-5-yl)ethanone	-	−5.16
1553	1-((S)-2,3-dihydro-2-((S)-1-hydroxypropan-2-yl)benzofuran-5-yl)ethanone	-	−4.67
1556	d-cyclocapillarol	-	−5.12
1557	l-cyclocapillarol	-	−5.13
1628	Pyrophen	-	−5
1691	Ephedralone	-	−4.35
1944	(+)-sydowic acid	-	−4.48
2570	Benzeneacetic acid, alpha-(hydroxymethyl)	-	−4.3
2572	p-hydroxycinnamic acid, ethyl ester	-	−4.2
2584	3(2h)-furanone, dihydro-2,2-dimethyl-5-phenyl	-	−3.6
2682	Flindersine	Asn516, Asn177, His554, Tyr515, Asn516	−4.6
2699	1-methyl-4-methoxy-2-quinolone	-	−3.43
2717	Argaminolic A	-	−4.42
2810	3-phenylacetoxy-6-hydroxytropane	-	−4.01
2817	6-hydroxyapoatropine	-	−4.08
2980	Bassiatin	-	−4.2
3045	Aspergillus diketopiperazine 7	-	−4.07
3046	Aspergillus diketopiperazine 8	-	−4.55
3076	Sulfimarin	-	−4.38
3084	1,2,3,11b-tetrahydroquinolactacide	-	−4.51
3088	Indoleacetic acid methyl ester	-	−3.74
3148	Isoarnottinin	Asn179, Ser486, Asn516, Asn516, Asn177, His554	−5.29
3155	Lomatin	-	−3.05
4143	Ethyl caffeate	-	−4.33
4253	p-methoxyacetophenone	-	−3.94
4466	6-acetyl-2,2-dimethylchromene	-	−4.27
4558	Methyl salicylate	-	−4.1

Docking Studies

The most similar 40 natural African metabolites were docked using the MOE software, with the co-crystallized ligand VXG serving as a reference compound. The fundamental purpose of this research was to know more about the orientations and binding modes of the tested compounds at the SARS-CoV-2 helicase enzyme's binding site (PDB ID: 5RMM). The docking method was validated by redocking the co-crystallized ligand in the active site of the SARS-CoV-2 helicase enzyme using the docking algorithm. The validation step established the protocol's applicability for the intended docking study as demonstrated by the small RMSD (1.15 Å) between the redocked pose and the co-crystallized one. Moreover, the ability of the docking algorithm to retrieve the co-crystallized ligand's reported binding mode also indicated the validity of the selected docking algorithm (Figure S2, in the Supplementary data).

The binding free energy (ΔG) between the docked molecules and the active site, as well as the correct binding mode, were considered in the comparison between the tested compounds. Table 2 shows the calculated ΔG (binding free energies) of the tested compounds and the reference molecule (VXG) against the SARS-CoV-2 helicase enzyme. It is noteworthy that the docking results showed that compound 3148 illustrated high-affinity values of −5.29 kcal/mol, which is higher than the redocked ligand VXG (−5.24 kcal/mol). Furthermore, binding energy scores of −5.00 to −5.16 kcal/mol were found in compounds 1628, 1552, 1556, 1557, and 1549, which are comparable to the redocked ligand VXG. The other compounds, on the other hand, had lower affinity values than the ligand VXG.

The redocked ligand VXG predicted binding mode revealed an affinity value of −5.24 kcal/mol. The formation of three hydrogen bond interactions is most likely responsible for the high binding affinity. One was formed between the C=O group of the carboxylate moiety and Ser486, while the other two were formed with Asn516 and Asn177. Additionally, the nitrogen atom of the pyrrolidine ring formed hydrophobic interactions (pi–cation interactions with Tyr269 and Tyr265. These findings were deemed to be in alignment with the reported data²⁴ (Figures 2 and S3).

Figure 2.

2D binding mode of ligand (VXG) in the active site of SARS-CoV-2 helicase.

With a docking score of −4.55 kcal/mol, compound 1165 fitted well into the active site of the SARS-CoV-2 helicase enzyme and formed four hydrogen bond interactions. The carbonyl group formed three hydrogen bonds with the essential amino acids Asn516 and Asn1177. Moreover, the OH group of the carboxylate group formed one hydrogen bond with Ser486. Additionally, the phenyl ring formed a hydrophobic interaction with Pro408 (Figures 3A and S4).

Figure 3.

2D binding modes of the tested compound in the active site of SARS-CoV-2 helicase. (A) Compound 1165, (B) compound 2682, (C) compound 1552, and (D) compound 3148.

Compound 2682 had a significant potential binding affinity (ΔG = −4.60 kcal/mol) into the SARS-CoV-2 helicase enzyme active site, as shown in Figures 3B and S5. The creation of two hydrogen bonds with the crucial amino acids Asn516 and Asn177 was thought to be responsible for the high binding affinity. In addition to the formation of a hydrophobic interaction (pi–pi bonds) with His554 and Tyr515, as well as one pi–cation interaction with Asn516.

The binding mode of compound 1552 (affinity value of −5.16 kcal/mol), which is extremely close to ligand VXG, revealed that the oxygen atom of the tetrahydrofuran moiety formed two hydrogen bonds with fundamental amino acids Asn534 and Asn177. In addition, the OH group formed another two hydrogen bond interactions with the same amino acids, Asn534 and Asn177. Furthermore, it formed a hydrophobic interaction with Asp534 and His554 (Figures 3C and S6).

The establishment of five hydrogen bonds with the fundamental amino acids in the active site of the SARS-CoV-2 helicase enzyme was shown by the binding mode of compound 3148 (affinity value of −5.29 kcal/mol), which was better than the ligand VXG. One was formed between the OH group and Asn179. There were also two hydrogen bonds between the C=O group and Ser486 and Asn516, as well as another two hydrogen bond interactions between the oxygen atom of the chromene ring and Asn516 and Asn177. Furthermore, it formed aromatic stacking interactions (two pi–pi bonds) with His554 (Figures 3D and S7).

ADMET Studies

The approval of any compound to be a drug is based on its pharmacokinetic assessment besides the biological activity. Consequently, the analysis of the ADMET properties of any new compound should be investigated in an early stage of drug discovery to avoid late drug withdrawal.⁵⁰

Fifteen compounds (Table 3) showed high binding energy with accepted binding mode against the prospective target (SARS-CoV-2 helicase). These compounds were subjected to ADMET studies. ADMET profile identifies absorption, distribution, metabolism, excretion, besides toxicity. Although the presence of varied in vitro studies that can investigate the ADMET properties, the in silico studies are more advantageous due to the limitations of cost, time, effort in addition to the strict regulations regarding animal lives.⁵¹ The predicted ADMET profiles of the 15 compounds were shown in Table 3 and Figure 4. The applied descriptors were (1) blood–brain barrier penetration (BBB), (2) absorption level from the human intestine (HA), (3) aqueous solubility level (S), (4) CYP2D6 binding and inhibition, and (5) plasma protein binding ability (PPB). Fortunately, the ADMET profiles of all compounds indicated their likeness to be drugs.

Figure 4.

The expected ADMET study. ADMET_AlogP98: lipid-water partition coefficient; ADMET_PSA_2D: polar molecular surface area. Two-dimensional polar surface area (PSA_2D) each drug is plotted against their computed atom-type partition coefficient (ALogP98). The area encompassed by the ellipse represents good absorption without any violation of the ADMET properties. Based on Egan et al,⁵² absorption model the 95% and 99% confidence limit ellipses corresponding to the blood–brain barrier (BBB) and intestinal absorption models are indicated.

Table 3.

Predicted ADMET Descriptors.

Comp.	BBB^a	HA^b	S^c	CYP2D6^d	PPB^e
Remdesivir	4	3	3	F	F
956	3	0	4	F	F
1165	2	0	4	F	T
1166	3	0	4	F	T
1337	2	0	3	F	T
1546	2	0	4	F	F
1549	2	0	4	F	T
1550	2	0	4	F	T
1552	2	0	3	F	T
1553	2	0	3	F	T
1556	2	0	3	F	T
1557	2	0	3	F	T
1628	3	0	3	F	F
2682	2	0	3	F	T
3046	3	0	4	F	F
3148	2	0	3	F	T

^a BBB: 0 is very high, 1 is high, 2 is medium, 3 is low, and 4 is very low.

^b HA: 0 is good, 1 is moderate, 2 is poor, and 3 is very poor.

^c S: solubility level, 0 is extremely low, 1 is very low, 2 is low, 3 is good, and 4 is optimal.

^d CYP2D6: F is a noninhibitor and T is an inhibitor.

^e PPB: F means less than 90% and T means more than 90%.

Toxicity Studies

The toxicity estimation in the early stage is of the utmost importance to minimize the failure of the drug in the late development or the clinical stages.⁵³ The utilization of in silico approaches in toxicity prediction played an indispensable role in drug development to avoid ethical regulations and resources availability besides time wasting in the usual in vitro and in vivo studies.⁵⁴ The basic principle of in silico toxicity prediction is the SAR-predictive toxicity. In detail, the computer compares the basic chemical structural properties of the examined molecules against the structures of thousands of compounds of reported safety and or toxicity⁵⁵ (see the detailed toxicity report of the examined compounds in the Supporting data). We employed Discovery Studio software to compute the toxicity profiles of the selected molecules against seven different models. The employed models were FDA rat carcinogenicity in mouse-female (FDA),^56,57 mouse carcinogenic potency (median toxic dose [TD₅₀]),⁵⁸ rat maximum tolerated dose (MTD),^59,60 rat oral median lethal dose (LD₅₀),⁶¹ rat chronic lowest observed adverse effect level (LOAEL),^62,63 ocular irritancy, and skin irritancy.⁶⁴ The obtained results (Table 4) eliminated compounds 1337, 1546, 2682, and 3148 as they computed to have carcinogenic potencies.

Table 4.

Toxicity Properties of Tested Compounds and the Reference.

Comp.	FDA^a	TD₅₀ (mg/kg/day)	MTD (g/kg)	LD₅₀ (g/kg)	LOAEL (g/kg)	Ocular irritancy^b	Skin irritancy^b
Remdesivir	N	9.246	0.235	0.309	0.004	M	M
956	N	739.158	0.756	1.658	0.161	D	N
1165	N	506.246	0.351	1.19	0.144	N	M
1166	N	876.555	0.394	2.297	0.061	M	N
1337	C	188.182	0.06	0.935	0.091	N	M
1546	C	400.718	0.486	3.308	0.258	S	M
1549	N	401.787	0.558	1.653	0.136	S	N
1550	N	401.787	0.558	1.653	0.136	S	N
1552	N	133.466	0.098	0.472	0.037	S	M
1553	N	133.466	0.098	0.472	0.037	S	M
1556	N	137.178	0.06	0.871	0.007	M	M
1557	N	137.178	0.06	0.871	0.007	M	M
1628	N	400.854	0.037	0.344	0.018	M	N
2682	C	105.286	0.057	4.113	0.033	M	N
3046	N	107.144	0.223	1.145	0.109	D	N
3148	C	125.319	0.357	0.893	0.06	S	M

^aN, not a carcinogen; C, carcinogen.

^bN, nonirritant; M, mild; D, moderate; S, severe.

Molecular Dynamic Simulation

While molecular docking predicts ligand binding, it has limitations in studying enzyme rigidity.⁶⁵ MD simulations, on the other hand, offer detailed insights into atomic-level protein behavior and accurate temporal resolution,⁶⁶ capturing conformational changes and energy alterations in the enzyme–ligand complex.⁶⁷ MD simulations compute forces on each atom in the ligand, protein, and complex at ultra-short intervals using a force field, reflecting potential energy.⁶⁸ Integration follows Newton's motion laws, considering bonded and nonbonded interactions.⁶⁹ This atomic-level calculation, repeated billions of times, yields atomic trajectories for a defined time interval.⁷⁰

Compound 1552, 1-((S)-2,3-dihydro-2-((R)-1-hydroxypropan-2-yl)benzofuran-5-yl)ethenone, has been favored to undergo MD simulation (for 200 ns) studies against SARS-CoV-2 helicase (PDB ID: 5RMM) as it exhibited the best binding in the docking studies and was compared with the reference compound (VXG). After the first 20 ns, the RMSD of the helicase in the two systems is, on average, around 3 Å (Figure 5A) indicating stable conformation. Comparing the RMSD of the two compounds (Figure 5B1), we can observe that compound 1552's RMSD exhibits significant values throughout the simulation, averaging approximately around 15 Å. The reason for these large fluctuations in compound 1552's RMSD can be shown by comparing the ligand conformation at 10 ns (green sticks), 27 ns (cyan sticks), 68 ns (magenta sticks), and 107 ns (yellow sticks) with the starting conformation (white sticks) in Figure 5B2. As can be seen, this increase is the result of a change in the location of the ligand with respect to the protein and its conformational change.

Figure 5.

Shows the (A) helicase RMSD graph, (B1) shows the compound 1552 and VXG RMSD graph, (B2) shows a comparison between compound 1552 conformation at 10 ns (green sticks), 27 ns (cyan sticks), 68 ns (magenta sticks), and 107 ns (yellow sticks) with the starting conformation (white sticks).

Additionally, Figure 6A the radius of gyration (RoG) and Figure 6B solvent-accessible surface area (SASA) show that, on average, the two systems exhibit very similar behavior over the whole simulation. From what can be seen in Figure 6C, the average number of H-bonds formed between each ligand and helicase systems shows different values. While VXG shows mainly two H-bonds, compound 1552 shows nearly one H-bonds throughout the simulation. Also, the RMSF of the C-alpha atoms in the helicase protein of both systems shows the same trend (Figure 6D). Compound 1552 shows slightly larger fluctuations in its separation between its center of mass and the helicase center of mass with an average of 12.5 Å compared to the distance between the protein and VXG that shows an average of 11.5 Å (calculated after excluding the first 65 ns) (Figure 6E).

Figure 6.

(A) Helicase protein’s RoG graph, (B) helicase protein’s SASA graph, (C) change in the hydrogen bonds numbers between each ligand and helicase, (D) helicase protein's RMSF graph, (E) distance from the center of mass of compound 1552 or VXG and helicase protein.

Molecular Mechanics-Generalized Boltzmann Surface Area

The MM-GBSA method was employed to calculate the free binding energy of the SARS-CoV-2 helicase-1552 complex. The MM-GBSA is an advanced method that can examine the binding between a protein and a specific ligand by computing the exact binding free energy of the ligand–protein complex over a determined time.⁷¹

The different parts of the MM-GBSA estimated binding free energy are shown in Figure 7. Considering that compound 1552 has a binding energy of −14.21 kcal/mol, we may infer that its interaction with the protein is moderate. Binding stability seems to be influenced more by Van der Waals interactions (−22.55 kcal/mol) than electrostatic ones (−12.47 kcal/mol). After that, we used decomposition analysis to quantify the contribution of amino acids located within 1 nm of compound 1552 (Figure 8). Amino acids that contribute more than −0.5 kcal/mol include Arg178 (−0.88 kcal/mol), Met378 (−0.58 kcal/mol), and Gln537 (−1.03 kcal/mol).

Figure 7.

MM-GBSA energy constituents and their values. Bars represent the standard deviations.

Figure 8.

Analysis of SARS-CoV-2 helicase-1552 complex binding free energy.

Protein–Ligand Interaction Fingerprint Study

Protein–ligand interaction fingerprint (ProLIF) is a crucial technique in the realm of computer-aided drug design, molecular docking, and MD investigations. It plays a central role in thoroughly examining and characterizing the interactions that occur between proteins and ligands. This method involves the creation of interaction fingerprints, which are unique patterns resulting from the interactions between a protein and a ligand. These fingerprints are vital for quantifying both the strength and the nature of these binding interactions.⁷² ProLIF enables the quantification of various types of interactions, including significant categories like hydrogen bonds, hydrophobic contacts, and other noncovalent associations. Furthermore, when applied in MD simulations, ProLIF serves as an invaluable tool for tracking the dynamic behaviors of protein–ligand complexes over extended periods. It provides crucial insights into how these interactions between the protein and ligand evolve throughout the simulation, thereby enhancing our comprehension of complex stability and binding affinity.⁷³ Results from the ProLIF library showed that three amino acids had an incidence of hydrophobic interaction greater than 73%. The three most common amino acids found to interact with compound 1552 through hydrophobic interactions are Ala312 (89.2%), Gln537 (85.8%), and Gly538 (73.4%) (Figure 9).

Figure 9.

ProLIF python library output: the interacted helicase amino acids with compound 1552, and their types of interactions as well as occurrence.

Protein–Ligand Interaction Profiles Study

PLIP, a prominent bioinformatics approach, plays a crucial role in dissecting and illustrating interactions occurring within molecular complexes involving a protein and a ligand. This computational tool offers invaluable insights into the binding interactions and noncovalent bonds established between the ligand and the protein, shedding light on the intricate molecular mechanisms governing ligand–receptor interactions.⁷⁴ Given its paramount relevance across diverse domains, including drug discovery, computational biology, and structural bioinformatics, PLIP has gained widespread adoption for the comprehensive examination of protein–ligand complexes and their interactions. When integrated with other computational methods such as molecular docking, MD simulations, and free energy calculations, PLIP enables a profound understanding of ligand–protein interactions.⁷⁵ Consequently, PLIP stands as a pivotal method in the realm of rational drug design. In our study, we utilized PLIP to extract 3D binding interactions in .pse file format from the representative frames obtained through clustering.⁷⁶ From the clustering-generated representative frames, the 3D binding interactions were extracted as.pse files using PLIP (Figure 10).

Figure 10.

The six representative clusters obtained from TTClust analysis of compound 1552 and the helicase protein, along with their corresponding time. Hydrophobic interactions are denoted by gray dashed lines, hydrogen bonds by blue solid lines, compound 1552 is represented by orange sticks, and amino acids of the helicase protein are shown as blue sticks.

Principal Component Analysis of Trajectories Studies

PCAT, a well-established computational technique widely employed in the realm of MD simulations, plays a pivotal role in the analysis and extraction of essential collective motions from trajectory data. During an MD simulation, precise records of atomic positions within the complex biomolecular system are continuously logged over time, resulting in a comprehensive trajectory that intricately captures the system's dynamic behavior.⁷⁷ At this juncture, PCAT's intervention becomes indispensable, as it simplifies this intricate trajectory. This simplification is achieved by projecting the trajectory onto a lower-dimensional space while preserving the most critical motions. By orchestrating this simplification, PCAT effectively highlights the dominant motions governing the system's behavior and functionality, providing invaluable insights into the fundamental collective motions that underlie the biomolecular system's dynamics. This coordinated motion analysis was performed using principal component analysis (PCA).^78,79

PCAT was utilized to identify the coordinated motions of the systems. According to the “Methods” section, we used the scree plot, the eigenvector distribution, and the variance preserved by each additional eigenvector (cumulative total) to determine the appropriate size of the reduced subspace. According to the scree plot, the slope seems to flatten out at the third PC. Additionally, the first eigenvector alone contained around 65.5% of the variation, while the first three eigenvectors together maintained over 79% of the overall variance (Figure 11). The first four principal components were shown to have a non-Gaussian distribution (Figure 12). Therefore, we chose the top three eigenvectors to determine the essential subspace.

Figure 11.

Eigenvalues variation with increasing the eigenvectors (blue) with the cumulative variance retained in the eigenvectors represented (red).

Figure 12.

The distribution of the first 10 eigenvectors.

Both the apo helicase and holo helicase simulations had their cosine content calculated, and the randomness of the first 10 eigenvectors was evaluated. In both the apo helicase and holo helicase proteins, the cosine content of the first 10 eigenvectors was below 0.2 (Figure 13). The similarity between the two subspaces (the first three eigenvectors) is 34.6% according to the Root Mean Square Inner Product (RMSIP), showing that the two trajectories are more or less comparable. In addition, the RMSIP indicated that the C matrices are 48.7% identical proving the similarity between the two trajectories.

Figure 13.

Cosine content of the first 10 eigenvectors for the two (apo helicase and holo helicase proteins) trajectories.

Figures 14--16 analysis of the results obtained of projecting each trajectory onto the first three eigenvectors of the new C matrix. Larger dots in these graphs indicate average structures for the indicated trajectories. Figure 14, a projection on the first two eigenvectors, reveals that the two trajectories have separate average structures and that the frames represent unique sampling with just a little amount of overlap before (pale red and white dots) and throughout the simulation (gray and orange dots). As can be seen in Figure 15, the two trajectories show approximate complete overlap and have roughly similar average structures. Figure 16 displays the projection on the second and third eigenvectors, which shows a similar trend to the first projection with just a little amount of overlap between the two trajectories.

Figure 14.

First two eigenvectors projection.

Figure 15.

First and third eigenvectors projection.

Figure 16.

Second and third eigenvectors projection.

The results obtained from the study are promising and showcase significant potential in the drug discovery of anti-COVID-19 treatments. However, it is imperative to acknowledge that the study primarily relies on in silico methodologies, involving computational simulations and predictions. While these computational techniques provide valuable insights and serve as a crucial initial step, their outcomes necessitate validation through actual experimental confirmation. Recognizing this, it becomes crucial to translate the computational predictions into tangible experimental evidence to ensure the reliability and applicability of the identified compounds especially compound 1552. Further experimental confirmation through in vitro and in vivo studies is essential to validate the efficacy, safety, and potential therapeutic applications of the promising compounds identified in the in silico analysis. This iterative process of computational predictions followed by rigorous experimental validation is integral to establishing the credibility and translational potential of the study's findings in the broader context of drug development.

Method

Molecular Similarity Detection

The molecular similarity of the 4924 natural African metabolites was assessed using Discovery Studio 4.0, with VXG as the reference molecule to find the most similar 200 compounds among them to the VXG. The analysis included quantitative parameters such as the number of rotatable bonds, rings, aromatic rings, hydrogen bond donors, hydrogen bond acceptors, ALog p, molecular weight, and MFPSA. Following compound preparation, the small molecules panel with library analysis was employed. Similar molecules were identified by numeric properties, using the prepared compounds as input ligands and VXG as the reference. The task was set to find the most similar molecules (30 in total) based on Euclidean distance. The output chart of structural similarity was visualized to interpret the results, as elucidated in greater detail within the Supplementary Data section.

Fingerprint Studies

The fingerprints of the tested 200 compounds, with VXG as the reference molecule, were assessed using Discovery Studio 4.0. These fingerprints were based on parameters such as the number of charges, hybridization, hydrogen bond acceptors, hydrogen bond donors, positive and negative ionizable atoms, halogen atoms, aromatic rings, and ALogP (octanol–water partition coefficient). The protocol involved preparing the compounds, activating the small molecules panel, enabling the library analysis option, and selecting similar molecules by fingerprints. The input ligands were the prepared compounds and VXG as the reference. The task was set to find the most similar molecules (15 in total) using the Tanimoto coefficient, and the visualized output provided the results of the protocol as elucidated in greater detail within the Supplementary Data section.

Docking Studies

The SARS-CoV-2 helicase crystal structure (PDB ID: 5RMM) was prepared for docking by removing water molecules and retaining one protein chain with the co-crystallized ligand VXG. Protonation and energy minimization were performed using MOE 2019 and MMFF94x, defining the active site based on residues within 5 Å of VXG. 2D structures of compounds and VXG were converted to 3D structures after protonation and energy minimization using MM2 force field. MOE 2019 conducted protein–ligand docking, validated through VXG redocking. The accuracy of the protocol was confirmed by comparing the redocked VXG RMSD with the co-crystallized reference. For tested compounds, 30 docked solutions were generated, and Discovery Studio 4.0 visualized and analyzed results against the VXG reference⁸⁰ as elucidated in greater detail within the Supplementary Data section.

ADMET Studies

ADMET descriptors for compound evaluation were determined using Discovery Studio 4.0. The CHARMM force field was initially applied, followed by the preparation and minimization of the compounds. Models covered human intestinal absorption, aqueous solubility, blood–brain barrier penetration, plasma protein binding, CYP2D6 inhibition, and hepatotoxicity. Molecules were filtered based on SMARTS® rules. After compound preparation, the small molecules panel activated the ADMET descriptors option. Prepared compounds were selected as input ligands, and all relevant ADMET parameters were chosen. The protocol output was visualized, generating an interpretable ADMET chart⁸¹ as elucidated in greater detail within the Supplementary Data section.

Toxicity Studies

Discovery Studio 4.0 was utilized for computing toxicity parameters of selected compounds, followed by a comprehensive evaluation using the Toxicity Prediction (Extensible) Protocol (TOPKAT). TOPKAT employs validated quantitative structure toxicity relationship models to assess the compounds’ toxicity based solely on molecular structure. After compound preparation, the small molecules panel activated the toxicity prediction (extensible) option. Prepared compounds served as input ligands, and various toxicity models were selected. The similarity search task was enabled, and the detailed report task was set to generate a PDF file. The protocol output was visualized, presenting a concise toxicity report⁸² as elucidated in greater detail within the Supplementary Data section.

Molecular Dynamics Simulations

In this study, we conducted a 200 nanoseconds (ns) standard unbiased MD simulation using GROMACS 2021. The aim was to assess the stability of helicase alone, the Helicase_VXG complex, and the Helicase_Compound 1552 complex while examining structural changes between the two complexes. Input files for GROMACS were generated via the CHARMM-GUI server's solution builder module. The solvated system, comprising both the docked complex and apoprotein, occupied an 11.8 nm cube using the TIP3P water model, with a 1 nm padding from the farthest atom. Charge neutralization was achieved by adding 0.154 M NaCl ions. The CHARMM36m force field defined amino acid parameters for the helicase protein, TIP3P water model, and ions. For compound 1552 and VXG, the CHARMM general force field (CGenFF) was employed for parameterization⁸³ as elucidated in greater detail within the Supplementary Data section.

Molecular Mechanics-Generalized Born Surface Area

The ligand's binding energy was calculated using the molecular mechanics-generalized Born surface area (MM-GBSA) method, implemented in the gmx_MMPBSA program. To assess the contribution of amino acids within a 1 nm proximity of the ligand, a decomposition analysis was performed.^14,15 The ionic strength was adjusted to 0.154 M, and the solvation method (igb) was set to 5. The internal dielectric constant was set to 1.0, and the external dielectric constant to 78.5⁸⁴ as elucidated in greater detail within the Supplementary Data section.

Principal Component Analysis of Trajectories

PCAT was applied to the mass-weighted covariance matrix (C) obtained from atomic positions in MD trajectories. This approach has the potential to reveal patterns of coordinated motion within the system. Specifically, PCA was employed to investigate the patterns of alpha carbons within the helicase. The analysis focused on amino acids spanning from Gln11 to Asp583, excluding terminal amino acids⁷⁷ as elucidated in greater detail within the Supplementary Data section.

Conclusion

In conclusion, our comprehensive CADD study, utilizing 4924 African natural metabolites, has yielded promising results as through a rigorous selection process, we identified 11 metabolites, including O-coumaric acid, trans-cinnamic acid, trans-ferulic acid, and others, demonstrating potent inhibitory activity against the critical SARS-CoV-2 helicase enzyme. Our extensive MD simulations and energy calculations confirmed the stable binding of the most promising candidate, 1-((S)-2,3-dihydro-2-((R)-1-hydroxypropan-2-yl)benzofuran-5-yl)ethenone, to the SARS-CoV-2 helicase with low energy and optimal dynamics. These results were provided by MD simulations, MM-GBSA, PLIP, ProLIF, and PCAT studies over 200 ns giving valuable insights into the molecular mechanisms underlying their inhibitory activity. These highly encouraging findings serve as a strong foundation for the next phase of our research, which will involve in vitro and in vivo studies to validate the efficacy and safety of these selected compounds. While more work lies ahead, our study raises hope for the development of potential COVID-19 therapeutics and contributes to the global effort to combat this pandemic.

Supplemental Material

sj-docx-1-npx-10.1177_1934578X241246738 - Supplemental material for Computer-Assisted Drug Discovery of Potential African Anti-SARS-CoV-2 Natural Products Targeting the Helicase Protein

Supplemental material, sj-docx-1-npx-10.1177_1934578X241246738 for Computer-Assisted Drug Discovery of Potential African Anti-SARS-CoV-2 Natural Products Targeting the Helicase Protein by Ahmed M. Metwaly, Mohamed S. Alesawy, Bshra A. Alsfouk, Ibrahim M. Ibrahim, Eslam B. Elkaeed and Ibrahim H. Eissa in Natural Product Communications

Footnotes

Acknowledgments

The authors would like to express sincere gratitude to AlMaarefa University, Riyadh, Saudi Arabia, for supporting this research.

Declaration of Conflicting Interests

All author(s) of this manuscript declare that they have no conflicts of interest to disclose. This includes, but is not limited to, any financial interests, professional relationships, or personal affiliations that might be perceived as influencing the research, analysis, or presentation of the findings contained within this work. The integrity and objectivity of this research remain paramount, and there are no competing interests that could compromise the impartiality of the study or its results.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by Princess Nourah bint Abdulrahman University Researchers Supporting Project number (PNURSP2024R142), Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia.

Ethical Approval

Not applicable for this article as no experiments or studies involving human or animal participants were conducted as part of this research.

Statement of Human and Animal Rights

This article does not include any research involving human or animal participants.

Statement of Informed Consent

Not available as no experiments or studies involving human or animal participants were conducted as part of this research.

Trial Registration Number/Date

Not available.

Grant Number

PNURSP2024R142.

ORCID iD

Ahmed M. Metwaly

Supplemental Material

Supplemental material for this article is available online.

References

WHO. WHO coronavirus (COVID-19) dashboard. Accessed September 10, 2021. https://covid19.who.int/

Hagler

. Chemoinformatics and drug discovery. Molecules. 2002;7(8):566-600.

Engel

. Basic overview of chemoinformatics. J Chem Inf Model. 2006;46(6):2267-2277. doi:https://doi.org/10.1021/ci600234z

Westbrook

Burley

. How structural biologists and the Protein Data Bank contributed to recent FDA new drug approvals. Structure. 2019;27(2):211-217.

Grimme

Schreiner

. Computational chemistry: the fate of current methods and future challenges. Angew Chem Int Ed Engl. 2018;57(16):4170-4176. doi:https://doi.org/10.1002/anie.201709943

Amin

Banerjee

Singh

Qureshi

Gayen

Jha

. First structure–activity relationship analysis of SARS-CoV-2 virus main protease (Mpro) inhibitors: an endeavor on COVID-19 drug discovery. Mol Diversity. 2021;25(3):1827-1838.

Idris

Yekeen

Alakanse

Durojaye

. Computer-aided screening for potential TMPRSS2 inhibitors: a combination of pharmacophore modeling, molecular docking and molecular dynamics simulation approaches. J Biomol Struct Dyn. 2021;39(15):5638-5656. doi:https://doi.org/10.1080/07391102.2020.1792346

. A new computer model for evaluating the selective binding affinity of phenylalkylamines to T-type Ca(2+) channels. Pharmaceuticals (Basel). 2021;14(2):141. doi:https://doi.org/10.3390/ph14020141

Fan

Zhang

. Progress in molecular docking. Quant Biol. 2019;7:83-89.

10.

Hospital

Goni

Orozco

Gelpi

. Molecular dynamics simulations: advances and applications. Adv Appl Bioinform Chem. 2015;8:37-47. doi:https://doi.org/10.2147/AABC.S70333

11.

Cheng

Liu

Tang

. In silico ADMET prediction: recent advances, current challenges and future trends. Curr Top Med Chem. 2013;13(11):1273-12-89. doi:https://doi.org/10.2174/15680266113139990033

12.

Obot

Macdonald

Gasem

. Density functional theory (DFT) as a powerful tool for designing new organic corrosion inhibitors. Part 1: an overview. Corros Sci. 2015;99:1-30.

13.

Ranjan

Devarapalli

Kundu

, et al. Isomorphism: ‘molecular similarity to crystal structure similarity’ in multicomponent forms of analgesic drugs tolfenamic and mefenamic acid. IUCrJ. 2020;7(Pt 2):173-183. doi:https://doi.org/10.1107/S205225251901604X

14.

Baidya

Ghosh

Amin

, et al. In silico modelling, identification of crucial molecular fingerprints, and prediction of new possible substrates of human organic cationic transporters 1 and 2. New J Chem. 2020;44(10):4129-4143.

15.

Metwaly

Lianlian

Luqi

Deqiang

. Black ginseng and its saponins: preparation, phytochemistry and pharmacological effects. Molecules (Basel, Switzerland). 2019;24(10):1856. doi:https://doi.org/10.3390/molecules24101856

16.

Metwaly

Ghoneim

Eissa

, et al. Traditional ancient Egyptian medicine: a review. Saudi J Biol Sci. 2021;28(10):5823-5832. doi:https://doi.org/10.1016/j.sjbs.2021.06.044

17.

Elkaeed

Youssef

Eissa

, et al. Multi-step in silico discovery of natural drugs against COVID-19 targeting main protease. Int J Mol Sci. 2022;23(13):6912. doi:https://doi.org/10.3390/ijms23136912

18.

Elkaeed

Eissa

Elkady

, et al. A multistage in silico study of natural potential inhibitors targeting SARS-CoV-2 main protease. Int J Mol Sci. 2022;23(15):8407.

19.

Elkaeed

Metwaly

Alesawy

Saleh

Alsfouk

Eissa

. Discovery of potential SARS-CoV-2 papain-like protease natural inhibitors employing a multi-phase in silico approach. Life (Basel). 2022;12(9):1407. doi:https://doi.org/10.3390/life12091407

20.

Elkaeed

Elkady

Belal

, et al. Multi-phase in silico discovery of potential SARS-CoV-2 RNA-dependent RNA polymerase inhibitors among 3009 clinical and FDA-approved related drugs. Processes. 2022;10(3):530.

21.

Eissa

Alesawy

Saleh

, et al. Ligand and structure-based in silico determination of the most promising SARS-CoV-2 nsp16-nsp10 2′-o-methyltransferase complex inhibitors among 3009 FDA approved drugs. Molecules. 2022;27(7):2287.

22.

Metwaly

Elwan

El-Attar

A-AMM

Al-Rashood

Eissa

Singh

. Structure-based virtual screening, docking, ADMET, molecular dynamics, and MM-PBSA calculations for the discovery of potential natural SARS-CoV-2 helicase inhibitors from the traditional Chinese medicine. J Chem. 2022;2022:1-23. doi:https://doi.org/10.1155/2022/7270094

23.

ANPDB. African natural products database; 2021 Accessed November 11, 2021. http://african-compounds.org/anpdb/compounds_list/

24.

Fearon

Powell

Douangamath

, et al. PanDDA analysis group deposition—crystal structure of SARS-CoV-2 main protease in complex with Z1220452176. RCSB Protein Data Bank. 2020.

25.

Hassell

Bledsoe

, et al. Crystallization of protein–ligand complexes. Acta Crystallogr D: Biol Crystallogr. 2007;63(1):72-79.

26.

Ripphausen

Nisius

Bajorath

. State-of-the-art in ligand-based virtual screening. Drug Discov Today. 2011;16(9-10):372-376.

27.

Burke

. Developments in molecular shape analysis to establish spatial similarity among flexible molecules. University of Illinois at Chicago, Health Sciences Center; 1993.

28.

Willett

. Similarity-based virtual screening using 2D fingerprints. Drug Discov Today. 2006;11(23-24):1046-1053.

29.

Briem

Kuntz

. Molecular similarity based on DOCK-generated fingerprints. J Med Chem. 1996;39(17):3401-3408.

30.

Willett

. Similarity searching using 2D structural fingerprints. Chemoinf Comput Chem Biol. 2010;672:133-158.

31.

Spackman

McKinnon

. Fingerprinting intermolecular interactions in molecular crystals. CrystEngComm. 2002;4(66):378-392.

32.

Chu

Q-X

Wang

Y-Q

Lin

Z-H

. In silico design of novel benzohydroxamate-based compounds as inhibitors of histone deacetylase 6 based on 3D-QSAR, molecular docking, and molecular dynamics simulations. New J Chem. 2020;44(48):21201-21210.

33.

Ieritano

Campbell

Hopkins

. Predicting differential ion mobility behaviour in silico using machine learning. Analyst. 2021;146(15):4737-4743.

34.

Taha

Ismail

Ali

, et al. Molecular hybridization conceded exceptionally potent quinolinyl-oxadiazole hybrids through phenyl linked thiosemicarbazide antileishmanial scaffolds: in silico validation and SAR studies. Bioorg Chem. 2017;71:192-200.

35.

Heikamp

Bajorath

. How do 2D fingerprints detect structurally diverse active compounds? Revealing compound subset-specific fingerprint features through systematic selection. J Chem Inf Model. 2011;51(9):2254-2265.

36.

Opo

FADM

Rahman

Ahammad

Ahmed

Bhuiyan

Asiri

. Structure based pharmacophore modeling, virtual screening, molecular docking and ADMET approaches for identification of natural anti-cancer agents targeting XIAP protein. Sci Rep. 2021;11(1):1-17.

37.

Duan

Dixon

Lowrie

Sherman

. Analysis and comparison of 2D fingerprints: insights into database screening performance using eight fingerprint methods. J Mol Graphics Modell. 2010;29(2):157-170.

38.

Sastry

Lowrie

Dixon

Sherman

. Large-scale systematic analysis of 2D fingerprint methods and parameters to improve virtual screening enrichments. J Chem Inf Model. 2010;50(5):771-784.

39.

Kogej

Engkvist

Blomberg

Muresan

. Multifingerprint based similarity searches for targeted class compound selection. J Chem Inf Model. 2006;46(3):1201-1213.

40.

Muegge

Mukherjee

. An overview of molecular fingerprint similarity search in virtual screening. Expert Opin Drug Discovery. 2016;11(2):137-148.

41.

Maggiora

Vogt

Stumpfe

Bajorath

. Molecular similarity in medicinal chemistry: miniperspective. J Med Chem. 2014;57(8):3186-3204.

42.

Turchi

Cai

Lian

. An evaluation of in-silico methods for predicting solute partition in multiphase complex fluids – a case study of octanol/water partition coefficient. Chem Eng Sci. 2019;197:150-158.

43.

Sullivan

Enoch

Ezendam

Sewald

Roggen

Cochrane

. An adverse outcome pathway for sensitization of the respiratory tract by low-molecular-weight chemicals: building evidence to support the utility of in vitro and in silico methods in a regulatory context. Appl In Vitro Toxicol. 2017;3(3):213-226.

44.

Altamash

Amhamed

Aparicio

Atilhan

. Effect of hydrogen bond donors and acceptors on CO₂ absorption by deep eutectic solvents. Processes. 2020;8(12):1533.

45.

Wan

Tian

Wang

Liu

. In silico studies of diarylpyridine derivatives as novel HIV-1 NNRTIs using docking-based 3D-QSAR, molecular dynamics, and pharmacophore modeling approaches. RSC Adv. 2018;8(71):40529-40543.

46.

Escamilla-Gutiérrez

Ribas-Aparicio

Córdova-Espinoza

Castelán-Vega

. In silico strategies for modeling RNA aptamers and predicting binding sites of their molecular targets. Nucleosides Nucleotides Nucleic Acids. 2021;40(8):798-807.

47.

Kaushik

Kumar

Bharadwaj

Chaudhary

Sahi

. Ligand-based approach for in-silico drug designing. In: Bioinformatics techniques for drug discovery. Springer; 2018:11-19.

48.

Jain

. Morphological similarity: a 3D molecular similarity method correlated with protein-ligand recognition. J Comput-Aided Mol Des. 2000;14(2):199-213.

49.

Zhang

Ren

J-X

Ding

. Development of an in silico prediction model for chemical-induced urinary tract toxicity by using naïve Bayes classifier. Mol Diversity. 2019;23(2):381-392.

50.

Ferreira

Andricopulo

. ADMET modeling approaches in drug discovery. Drug Discov Today. 2019;24(5):1157-1165.

51.

Norinder

Bergström

. Prediction of ADMET properties. ChemMedChem: Chem Enabl Drug Discovery. 2006;1(9):920-937.

52.

Egan

Merz

Baldwin

. Prediction of drug absorption using multivariate statistics. J Med Chem. 2000;43(21):3867-3877.

53.

Dearden

. In silico prediction of drug toxicity. J Comput-Aided Mol Des. 2003;17(2):119-127.

54.

Idakwo

Luttrell

Chen

, et al. A review on machine learning methods for in silico toxicity prediction. J Environ Sci Health C. 2018;36(4):169-191.

55.

Kruhlak

Benz

Zhou

Colatsky

. (Q) SAR modeling and safety assessment in regulatory review. Clin Pharmacol Ther. 2012;91(3):529-534.

56.

Xia

Maliski

Gallant

Rogers

. Classification of kinase inhibitors using a Bayesian model. J Med Chem. 2004;47(18):4463-4470.

57.

BIOVIA. QSAR, ADMET and predictive toxicology. Accessed May, 2020, https://www.3dsbiovia.com/products/collaborative-science/biovia-discovery-studio/qsar-admet-and-predictive-toxicology.html

58.

Venkatapathy

Wang

NCY

Martin

Harten

Young

. Gen Appl Syst Toxicol. 2009

59.

Goodrnan

Wilson

. Comparison of the dependence of the TD50 on maximum tolerated dose for mutagens and nonmutagens. Risk Anal. 1992;12(4):525-533.

60.

National Research Council. Correlation between carcinogenic potency and the maximum tolerated dose: implications for risk assessment. In: Issues in risk assessment. National Academies Press (US); 1993.

61.

Gonella Diaza

Manganelli

Esposito

Roncaglioni

Manganaro

Benfenati

. Comparison of in silico tools for evaluating rat oral acute toxicity. SAR QSAR Environ Res. 2015;26(1):1-27.

62.

Pizzo

Benfenati

. In silico models for repeated-dose toxicity (RDT): prediction of the no observed adverse effect level (NOAEL) and lowest observed adverse effect level (LOAEL) for drugs. In: In silico methods for predicting drug toxicity. Springer; 2016:163-176.

63.

Venkatapathy

Moudgal

Bruce

. Assessment of the oral rat chronic lowest observed adverse effect level model in TOPKAT, a QSAR software package for toxicity prediction. J Chem Inf Comput Sci. 2004;44(5):1623-1629.

64.

Wilhelmus

. The Draize eye test. Surv Ophthalmol. 2001;45(6):493-515.

65.

Sousa

Fernandes

Ramos

. Protein–ligand docking: current status and future challenges. Proteins: Struct, Funct, Bioinf. 2006;65(1):15-26.

66.

Hollingsworth

Dror

. Molecular dynamics simulation for all. Neuron. 2018;99(6):1129-1143.

67.

Liu

Shi

Zhou

Liu

Yao

. Molecular dynamics simulations and novel drug discovery. Expert Opin Drug Discovery. 2018;13(1):23-37.

68.

Hansson

Oostenbrink

van Gunsteren

. Molecular dynamics simulations. Curr Opin Struct Biol. 2002;12(2):190-196.

69.

Durrant

McCammon

. Molecular dynamics simulations and drug discovery. BMC Biol. 2011;9(1):1-9.

70.

Kalyaanamoorthy

Chen

Y-PP

. Modelling and enhanced molecular dynamics to steer structure-based drug discovery. Prog Biophys Mol Biol. 2014;114(3):123-136.

71.

Homeyer

Gohlke

. Free energy calculations by the molecular mechanics Poisson−Boltzmann surface area method. Mol Inform. 2012;31(2):114-122.

72.

Xiong

Shen

Yang

, et al. Featurization strategies for protein–ligand interactions and their applications in scoring function development. Wiley Interdiscip Rev: Comput Mol Sci. 2022;12(2):e1567.

73.

Zhao

Bourne

. Harnessing systematic protein–ligand interaction fingerprints for drug discovery. Drug Discov Today. 2022;27:103319.

74.

Salentin

Adasme

Heinrich

, et al. From malaria to cancer: computational drug repositioning of amodiaquine using PLIP interaction patterns. Sci Rep. 2017;7(1):11401.

75.

Salentin

Schreiber

Haupt

Adasme

Schroeder

. PLIP: fully automated protein–ligand interaction profiler. Nucleic Acids Res. 2015;43(W1):W443-W447.

76.

Tubiana

Carvaillo

J-C

Boulard

Bressanelli

. TTClust: a versatile molecular simulation trajectory clustering program with graphical summaries. J Chem Inf Model. 2018;58(11):2178-2182.

77.

Papaleo

Mereghetti

Fantucci

Grandori

De Gioia

. Free-energy landscape, principal component analysis, and structural clustering to identify representative conformations from molecular dynamics simulations: the myoglobin case. J Mol Graphics Modell. 2009;27(8):889-899.

78.

David

Jacobs

. Principal component analysis: a method for determining the essential dynamics of proteins. Protein Dyn: Methods Protoc. 2014;1084:193-226.

79.

Maisuradze

Leitner

. Free energy landscape of a biomolecule in dihedral principal component space: sampling convergence and correspondence between structures and minima. Proteins: Struct, Funct, Bioinf. 2007;67(3):569-578.

80.

Suleimen

Jose

Mamytbekova

, et al. Isolation and in silico inhibitory potential against SARS-CoV-2 RNA polymerase of the rare kaempferol 3-O-(6 ″-O-acetyl)-glucoside from calligonum tetrapterum. Plants. 2022;11(15):2072.

81.

Biovia

. Discovery Studio Modeling Environment. Release; 2017.

82.

Yousef

Elwan

Gobaara

, et al. Anti-cancer and immunomodulatory evaluation of new nicotinamide derivatives as potential VEGFR-2 inhibitors and apoptosis inducers: in vitro and in silico studies. J Enzyme Inhib Med Chem. 2022;37(1):2206-2222.

83.

Cheng

Islam

, et al. CHARMM-GUI PDB manipulator for advanced modeling and simulations of proteins containing nonstandard residues. Adv Protein Chem Struct Biol. 2014;96:235-265.

84.

Valdés-Tresanco

Valiente

Moreno

. Gmx_MMPBSA: a new tool to perform end-state free energy calculations with GROMACS. J Chem Theory Comput. 2021;17(10):6281-6291.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.42 MB