Virtual screening of zinc compounds similar to NSAIDS with better pharmacodynamic and pharmacokinetic profiles

Ligands retrieval

The PubChem online tool was opened and aspirin, ibuprofen and celecoxib were searched. The summary button was clicked and this generated the smiles of the compounds. This also allowed the reference compounds (aspirin, ibuprofen and celecoxib) to be downloaded as SDF files. The downloaded files were saved to folders named aspirin, ibuprofen and celecoxib on the desktop of the laptop. Combined Zinc–Drug like was then clicked in SwissSimilarity, which automatically generated a red search button that was clicked on. Zinc compounds were generated.

Ligand-based virtual screening

Ligand-based virtual screening uses active ligands for lead identification and optimization rather than the structures of the protein. Using the SwissSimilarity online tool, the known compounds are used as queries in virtual screening to identify new active compounds. There must be similarities between the known ligands to the library of compounds in the form of geometrical measurements. Pharmacophore-based methods use molecular properties and then generate the similarity score between the known compounds and the ligand compounds. The molecular shape determines the overlap of shapes between the compounds and the known agents and creates the similarity scores. The assumptions of the molecular shape are that the overlap in structures of the known compounds and the ligands show activity (Hamza et al., 2012), therefore the ligands with different shapes to the query compounds are missed in virtual screening. The compounds are then scored and differentiated as active or inactive (Hamza et al., 2012).

In this study, aspirin, celecoxib, and ibuprofen were used as queries to generate zinc compounds based on the structures of the reference compounds (aspirin, celecoxib, and ibuprofen). The SwissSimilarity online tool was used. The SwissSimilarity tool generates agents with high drug-likeness to the agents by use of a vast database like ZINC (Bragina et al., 2022) and this gives a pool of structures to be analysed.

Inclusion criteria for ZINC compounds

ZINC compounds with above 80% similarity scores to the query compounds ibuprofen and aspirin were chosen, while compounds with a similarity score above 70% to celecoxib were chosen. For the compounds that had the same similarity scores, only one compound was included. 20 compounds for ibuprofen, 20 compounds for aspirin and 20 compounds for celecoxib were included giving a total of 60 compounds that were included in this study.

Structure-based virtual screening

Structure-based virtual screening starts with protein structure identification through NMR spectroscopy or homology modelling. This is followed by binding site identification which considers druggability, stereochemistry, and tautomers as well as protonation of the binding site. Receptor preparation is done by modifying the tautomeric forms of histidine receptors, removing unnecessary side chains, determining flexibility of the compounds as well as solvation of the receptor binding sites. Ligand preparation follows by ensuring they are at as minimal energy state as possible which is done by Chimera software (Pettersen et al., 2004). Docking of the ligands to the receptor is done using AutoDock Vina and scoring is done such that the higher the binding scores of the compounds, the higher the binding affinity (Eberhardt et al., 2021). Post docking analysis of the compounds is completed to show the 2D or 3D interactions of the ligands to their receptors, such as by using Discovery studio (RRID: SCR_015651) (Lionta et al., 2014).

In this study, cyclooxygenase 1 and 2 enzymes were prepared using Chimera (RRID: SCR_004097) software which eliminated the unwanted residues bound to the receptors. Ligand preparation was done using Avogadro software (RRID: SCR_011958). Molecular docking was done using AutoDock Vina (RRID: SCR_011958). Zinc compounds that showed higher docking scores in reference to aspirin, celecoxib and ibuprofen were further analyzed for pharmacokinetics as well as toxicity profiles.

Pharmacokinetics properties

The pharmacokinetics properties of the agents were analyzed using SwissADME, which lead to the discovery of agents with improved properties.

Toxicity predictions

Tox 21 (The Toxicology in the 21^st century) has led to the development of computational methods for the toxicological assessment of compounds from the identified toxicological pathways (Banerjee et al., 2018). ProTox II webserver is commonly used in toxicity predictions and predicts toxicities such as oral toxicity, organ toxicity (hepatoxicity), and toxicological endpoints which include mutagenicity, cytotoxicity, immunotoxicity, and immunogenicity. The molecular mechanism for predicting toxicities in the body is as well generated from the ProTox-II webserver (Banerjee et al., 2018).

ProTox II (RRID: SCR_018506) webserver was used in this study to determine toxicity profiles of zinc compounds and were compared to the reference compounds.

Data analysis

Descriptive analysis was chosen since the data involved was numerical. Compounds that showed better docking scores to COX-2 and lower docking scores to COX-1 than aspirin, ibuprofen and celecoxib were identified and were chosen as promising compounds. The highlighted compounds were determined for their toxicity profiles and pharmacokinetics properties. Pharmacokinetic data were analyzed non-numerically since the data generated was qualitative.

Docking results

Docking scores of aspirin and zinc compounds to Cox-1 and cox-2

Aspirin downloaded from PubChem was used as the reference against other zinc compounds. The ligands with better docking scores to COX-2 and lower docking scores to COX-1 in reference to aspirin are highlighted in Table 1 below (Faith et al., 2023). ZINC00001050; 0.995, ZINC00409844; 0.992, ZINC01651927; 0.991, ZINC01680731; 0.994, ZINC19405119; 0.986, ZINC33823423; 0.996, and ZINC71771212; 0.994, were analysed further for their pharmacokinetics properties using the SwissADME online webserver and toxicity predictions using the ProTox II webserver.

All the zinc and aspirin compounds showed high gastrointestinal (GIT) absorption and permeation to the blood-brain barrier (BBB). The compounds were shown not to be P-glycoprotein substrates and to lack enzyme inhibition properties to CYP (cytochrome) 1A2, CYP 2C19, CYP2C9, CYP2D6 and CYP3A4 enzymes. The compounds further comply with the Lipinski rule of five. Exceptions to this were ZINC01651927; 0.991 and ZINC19405119; 0.986, where ZINC01651927; 0.991 showed inhibition of CYP1A2 enzyme while ZINC19405119; 0.986 showed no permeation to the BBB.

Docking scores of ibuprofen and zinc compounds to COX-1 and COX-2

The reference compound (ibuprofen) had better docking scores to both COX-1 and COX-2 than the zinc compounds except ZINC39120409; 0.996, which showed a higher docking score to COX-2 than ibuprofen and a lower docking score to COX-1 than ibuprofen as illustrated in Table 2. ZINC39120409; 0.996 was further analyzed for its pharmacokinetic properties and toxicity profile.

ZINC39120409; 0.996 and ibuprofen showed high GIT absorption and permeation to the blood-brain barrier. The compounds shown not to be P-glycoprotein substrates and lack enzyme inhibition properties to CYP 1A2, CYP 2C19, CYP2C9, CYP2D6 and CYP3A4 enzymes. The compounds further complied with Lipinski rule of five.

Celecoxib and zinc compounds docking scores to COX-2 and 1

The highlighted compounds in Table 3 showed better docking scores to the COX-2 enzyme than celecoxib and lower docking scores to COX-1 than celecoxib. ZINC00600558; 0.988; ZINC34660892; 0.988; and ZINC43464372; 0.750; were further analyzed for their pharmacokinetics properties as well as toxicity parameters.

Celecoxib showed high GIT absorption and showed inhibition to CYP1A2 and CYP 2C9. Celecoxib lacks inhibition to CYP2C19, CYP2D6, and CYP3A4. Celecoxib is not a substrate to P-glycoprotein and lacks BBB permeation.

ZINC00600558; 0.988, ZINC34660892 and ZINC43464372; 0.750 showed high GIT permeability, and a lack of CYP2D6 and CYP3A4 inhibition. These zinc compounds are not substrates of P-glycoprotein, and they lack permeation to the BBB. ZINC00600558; 0.988 and ZINC43464372; 0.750 however, showed enzyme inhibition activity to CYP1A2, CYP2C19, and CYP2C9. ZINC34660892 showed a lack of enzyme inhibition to CYP1A2 and the presence of CYP2C9 and CYP2C19 inhibition. They all complied with Lipinski rule of five.

Toxicity

Oral toxicity

ZINC01680731; 0.994 and ZINC01680731; 0.994 were predicted to be the safest with oral LD50 of 1240 mg/kg as compared to aspirin which was predicted to have oral LD50 of 250 mg/kg. ZINC00409844; 0.992, ZINC01651927; 0.991 and ZINC19405119; 0.986 were predicted to have the same oral LD50 as aspirin of 250 mg/kg. ZINC00001050; 0.995 and ZINC71771212; 0.994 were predicted to have oral LD50 of 650mg/kg (see Table 4).

Ibuprofen and ZINC39120409; 0.996 were predicted to have similar oral LD50 of 299 mg/kg.

ZINC00600558; 0.988, ZINC34660892; 0.988 and ZINC43464372; 0.750 showed similar oral LD50 of 1400mg/g as celecoxib.

Hepatotoxicity and immunotoxicity

ZINC00001050; 0.995, ZINC71771212; 0.994 and ZINC33823423; 0.996 were predicted to be hepatotoxic with probability scores of 0.56, 0.56 and 0.55 respectively in relation to aspirin, which is inactive. Ibuprofen and ZINC39120409; 0.996 were predicted to be hepatotoxic with probability scores of 0.66 and 0.56 respectively. ZINC43464372; 0.750 was predicted to be hepatoactive with probability score of 0.58 in relation to celecoxib. None of the compounds was predicted to be immunotoxic (see Table 5).

Genotoxicity

Celecoxib, ZINC34660892; 0.988 and ZINC43464372; 0.750 were predicted to be carcinogenic with probability scores of 0.56, 0.59 and 0.67 respectively. The rest of the compounds were inactive for carcinogenic, mutagenic and cytogenic properties. ZINC00600558; 0.988 was predicted to be inactive for carcinogenesis as compared to celecoxib (see Table 6).

Pathway toxicity

ZINC00409844; 0.992 and ZINC01651927; 0.991 were predicted to have activity on aryl hydrocarbon receptors with probability scores of 0.5 as compared to aspirin, which was predicted to be inactive. Celecoxib, ZINC00600558; 0.988 and were predicted to be active for estrogen receptor with probability scores of 1.0, 0.7 and 0.93 respectively (see Table 7).

ZINC00409844; 0.992 and ZINC01651927; 0.991; were predicted to be active for mitochondrial membrane potential with a probability score of 0.67. ZINC43464372; 0.750 was also predicted to be active for mitochondrial membrane potential with a probability score of 0.52,relative to celecoxib that was predicted inactive (see Table 8).

Pharmacokinetics profile

BBB permeation

ZINC19405119; 0.986, celecoxib, ZINC00600558; 0.988, ZINC34660892; 0.988 and ZINC43464372; 0.750 lack blood brain barrier penetration while aspirin has blood brain barrier penetration (see Table 9).

Metabolism

ZINC01651927; 0.991, celecoxib, ZINC00600558; 0.988 and ZINC43464372; 0.750 were all predicted to be inhibitors of CYP1A2. ZINC00600558; 0.988, ZINC34660892; 0.988 and ZINC43464372; 0.750 were predicted to be inhibitors of both CYP2C9 and CYP2C19 in comparison to celecoxib, which lacks CYP2C9 and CYP2C19 inhibition (see Table 10).

Post-docking analysis (Faith et al., 2023)

Aspirin and ZINC01680731; 0.994 interactions with COX-2 receptor (Figure 1)

Figure 1. Aspirin and zinc compounds 2D-interactions with receptors.

In both aspirin and ZINC01680731; 0.994, the phenyl ring is involved in pi-pi binding with histidine at chain A of the receptor at position 388. The conventional hydrogen bond in aspirin is at tryptophan of chain A of position 387 while in ZINC01680731; 0.994, three conventional hydrogen bonds are at histidine at positions 207 of chain A, tryptophan at position 387 of chain A and threonine at position 206 of chain A. Van der Waals forces at positions 391 and 390 of leucine, 199 of alanine, 203 of glycine, and 206 and 385 of threonine in chain B of COX-2 receptor in aspirin are different from the van der Waals forces in ZINC01680731; 0.994 at positions 199, 202 and 385. These differences in binding of the compounds to the COX-2 receptor could have been attributed to ZINC01680731; 0.994 having a higher binding score compared to aspirin (Figure 1).

Ibuprofen and ZINC39120409; 0.996 2D-interactions with receptors (Figure 2)

Figure 2. Ibuprofen and zinc compounds 2D-interactions with receptors.

Both ibuprofen and ZINC39120409; 0.996 showed similar modes of interactions in COX-1 as compared to their binding to COX-2. Ibuprofen showed similar interactions to COX-2 as ZINC39120409; 0.996 with the exemptions that ibuprofen’s pharmacophore also bonded with pi-pi binding to active sites in COX-2 which was absent from ZINC39120409; 0.996 binding to COX-2 (Figure 2).

Celecoxib and its zinc compounds 2D-visualization (Figures 3 and 4)

Figure 3. Celecoxib 2D-interactions with receptors.

Figure 4. Zinc compounds’ 2D-interactions with receptors.

Celecoxib’s pharmacophore binds similarly to active sites in COX-1 and COX-2 with fluorine binding to active sites in both receptors (Figure 3). A halogen bond was noted in all the zinc compounds. Covalent bonding, van der Waals forces, alkyl, and pi-alkyl bonds were noted in all the compounds’ pharmacophores in binding with COX-1 active sites. Celecoxib’s pharmacophore interacted with active sites in COX-2 using van der Waals forces, covalent bonding, and pi-pi bonds, Pi-anion and pi-sigma bonds were noted in ZINC00600558; 0.988, and ZINC01028131; 0.765; respectively (Figure 4).