In-silico identification of COX-2 inhibitory phytochemicals from traditional medicinal plants: molecular docking, dynamics, and safety predictions

Abstract

Inflammation is an essential biological response that facilitates tissue repair and immune defense; however, chronic inflammation is associated with numerous pathological conditions, including cardiovascular diseases, autoimmune disorders, and cancer. Cyclooxygenase-2 (COX-2) is a key enzyme in this process, catalyzing the synthesis of pro-inflammatory prostaglandins, thus representing a critical target for anti-inflammatory therapies. Conventional COX-2 inhibitors, particularly non-steroidal anti-inflammatory drugs (NSAIDs), often have significant side effects, creating an urgent need for safer alternatives. This in-silico study evaluates the binding affinities of bioactive compounds from Gmelina arborea, Coriandrum sativum, Glycyrrhiza glabra, Terminalia chebula, Solanum nigrum, Vernonia cinerea, Portulaca oleracea, Azadirachta indica, and Thespesia populnea to the COX-2 receptor. Molecular docking and dynamics simulations identified solasonine, solamargine, rutin, and glycyrrhizin as having binding affinities ranging from − 9.40 to − 8.50 kcal/mol, exceeding that of the standard NSAID diclofenac (− 5.68 kcal/mol). While these docking results provide valuable insights, further in-vitro validation is necessary. Stability analysis of ligand-receptor complexes showed minimal structural fluctuations. Moreover, cardiotoxicity predictions indicated that solamargine, rutin, and glycyrrhizin present a lower risk compared to diclofenac. ADMET profiling highlighted favorable pharmacokinetic properties for rutin, suggesting its potential as a promising COX-2 inhibitor with a beneficial safety profile. Subsequent MM-GBSA calculations revealed binding free energy values of − 11.316 kcal/mol for rutin and − 35.190 kcal/mol for diclofenac, indicating strong binding interactions. Overall, this study underscores the potential of these natural compounds as safer alternatives in anti-inflammatory therapy, paving the way for future experimental validation and clinical application.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40203-025-00407-4.

Introduction

Inflammation is a critical biological response that serves as a protective mechanism against harmful stimuli, including pathogens, injuries, and irritants. This response plays an essential role in the immune system, facilitating tissue repair and defense (Ojo et al. 2023). Upon tissue damage or infection, the immune system activates a complex cascade involving various immune cells, signaling molecules, and vascular changes (Britzen-Laurent et al. 2023). Key players in this process are immune cells, particularly leukocytes, which migrate to the site of injury, releasing cytokines and chemokines that orchestrate the inflammatory response, promoting healing and restoring homeostasis (Medzhitov 2008; Kulkarni et al. 2016). Inflammation can be categorized into two primary forms: acute and chronic. Acute inflammation is characterized by a rapid onset and is marked by classic symptoms such as redness, swelling, warmth, and pain (Walker et al. 2023). This transient response aims to eliminate the initial cause of cellular damage and initiate the healing process (Serhan et al. 2008). In contrast, chronic inflammation is a prolonged response that can persist for weeks, months, or even years. It often arises when the immune system fails to eradicate the inciting agent, leading to ongoing tissue damage and dysregulation of inflammatory pathways (Caballero-Sánchez et al. 2024). Chronic inflammation is implicated in various pathological conditions, including autoimmune disorders (Abou-Raya et al. 2007), cardiovascular diseases (Manabe 2011; Lopez-Candales et al. 2017; Ferrucci and Fabbri 2018), and metabolic syndromes (Hotamisligil 2006; Monteiro and Azevedo 2010; Wu et al. 2014). While inflammation is crucial for host defense, excessive or dysregulated inflammatory responses can have detrimental effects on health. Chronic inflammation is associated with an increased risk of diseases such as cancer, diabetes, and neurodegenerative disorders (Coussens and Werb 2002). Therefore, effective management of inflammation is paramount. Traditional approaches to managing inflammation include lifestyle modifications, dietary interventions, and pharmacological treatments. Nonsteroidal anti-inflammatory drugs (NSAIDs), corticosteroids, and immunosuppressants are commonly employed to mitigate inflammation and alleviate associated symptoms (Rennie et al. 2003). However, the long-term use of these agents can lead to significant side effects, including gastrointestinal complications and cardiovascular risks (Yaksh et al. 1998). Cyclooxygenases (COX) are key enzymes in the inflammatory process (Khan et al. 2022), catalyzing the conversion of arachidonic acid to prostaglandins, which are bioactive lipid mediators (Kursun et al. 2022). Two isoforms have been identified: COX-1, constitutively expressed in most tissues, and COX-2, an inducible enzyme primarily activated at sites of inflammation (Mazaleuskaya and Ricciotti 2020). COX-2 is upregulated by pro-inflammatory stimuli, resulting in increased production of prostaglandins, particularly PGE2, which mediates pain, fever, and swelling (Funk 2001). Consequently, the therapeutic inhibition of COX-2 has become a focal point in developing anti-inflammatory agents (19–21). While NSAIDs effectively inhibit COX-2, their long-term use is associated with adverse cardiovascular events, underscoring the necessity for novel COX-2 inhibitors with improved safety profiles (21–23). In recent years, there has been a growing interest in natural products derived from medicinal plants, which have been utilized for centuries for their therapeutic properties. Traditional medicine systems such as Ayurveda, Unani, and Persian medicine have long recognized the value of these plants, often using them as primary sources for treating various ailments. These systems emphasize holistic approaches to health, integrating natural remedies that are rich in bioactive compounds. Such compounds may offer anti-inflammatory benefits with fewer side effects compared to synthetic pharmaceuticals (Newman and Cragg 2007; Naghizadeh et al. 2020; Irannejad 2023). Among the promising candidates are compounds derived from Gmelina arbore (Kaur et al. 2018), Coriandrum sativum, Glycyrrhiza glabra (El-Saber Batiha et al. 2020), Terminalia chebula, Solanum nigrum (Chen et al. 2022), Vernonia cinerea, Portulaca oleracea, Azadirachta indica, and Thespesia populnea, all of which are recognized for their anti-inflammatory properties.

Gmelina arborea, or Gamhar, belongs to the Verbenaceae family and is found in South Asia. Its leaves exhibit antioxidant (Vijay et al. 2011) and analgesic properties (Gangwar et al. 2013), while the bark shows antidiabetic (Pattanayak et al. 2011) and anti-inflammatory (Parhi et al. 2011) effects. In traditional practices, it has been used to treat various inflammatory conditions (Lawrence et al. 2016). Coriandrum sativum (coriander or cilantro) is an annual herb in the Apiaceae family, native to Europe, Asia, and North Africa. Known for its culinary uses, it also offers health benefits such as anti-inflammatory, antimicrobial, and antioxidant properties, and is commonly used in traditional medicine for digestive and skin issues (Nadeem et al. 2013; Agarwal et al. 2025; Ismail et al. 2025). Glycyrrhiza glabra Linn, or licorice, is a perennial herb from the Fabaceae family, native to the Mediterranean and parts of Asia. Its root has a longstanding history in traditional medicine for its anti-inflammatory, antiviral, and antioxidant properties, particularly in treating respiratory and digestive issues (Kaur et al. 2013; Yang and Wang 2015). Terminalia chebula, or Haritaki, is a medicinal tree native to India, Nepal, and Sri Lanka, renowned in Ayurvedic medicine for its digestive, antioxidant, anti-inflammatory, and antimicrobial properties (Gupta et al. 2010; Dodke and Pansare 2017). It is often referred to as the "king of medicines" in traditional texts, highlighting its significance in holistic health approaches (Sharma et al. 2019). Solanum nigrum Linn., or black nightshade, is a globally distributed member of the Solanaceae family. Valued in traditional medicine, it is noted for its anti-inflammatory, analgesic, and antipyretic effects, as well as its antioxidant, antimicrobial, and anticancer properties, underscoring its importance in herbal therapies (Tiwari 2023). Vernonia cinerea, or little ironweed, is a flowering plant in the Asteraceae family, native to tropical and subtropical regions worldwide. Traditionally, it has been used for its anti-inflammatory, antimicrobial, and digestive aid properties (Shelar et al. 2014; Dogra and Kumar 2015). Portulaca oleracea, or purslane, is a succulent annual plant in the Portulacaceae family. Rich in antioxidants, its leaves and stems are associated with several health benefits, notably anti-inflammatory and cardioprotective effects (Iranshahy et al. 2017; Kumar et al. 2022). Azadirachta indica, commonly known as neem, is a tree native to the Indian subcontinent, celebrated for its medicinal properties. Its leaves, bark, seeds, and oil are known for their antimicrobial, anti-inflammatory, and antidiabetic effects, making neem an important resource in traditional medicine and agriculture (Dubey and Kashyap 2014; Modi and Soni 2023; Mohanasundaram and Antoneyraj 2025). Thespesia populnea, or coastal mahoe, is a flowering plant in the Malvaceae family, native to tropical coastal regions of the Indian Ocean, Pacific Ocean, and the Caribbean. It is valued for its traditional medicinal uses, particularly in treating skin ailments and digestive issues (Pandiyan et al. 2024).

This study aims to investigate the anti-inflammatory effects of natural compounds derived from selected medicinal plants. Utilizing computational approaches such as molecular docking and molecular dynamics simulations, the research will analyze the interactions between these compounds and the COX-2 enzyme. By assessing their binding affinities, stability, and potential inhibitory effects at the active site of COX-2, this study seeks to identify promising natural alternatives to conventional NSAIDs (Fig. 1).

Fig. 1.

Research findings on selected plant compounds' anti-inflammatory effects via COX-2 modulation

Results

Molecular docking studies

To validate the docking parameters, the co-crystallized ligand was re-docked into the active site of the COX-2 protein (PDB ID: 1PXX). The results were consistent, reinforcing the reliability of the docking approach, as indicated by an RMSD value of 1.00 Å. Following this validation, molecular docking was employed to assess the interactions of nine plants: Gmelina arbore, Coriandrum sativum, Glycyrrhiza glabra, Terminalia chebula, Solanum nigrum, Vernonia cinerea, Portulaca oleracea, Azadirachta indica, and Thespesia populnea (Table 1), with the COX-2 protein (Kumar et al. 2022; Gu et al. 2018; Pastorino et al. 2018; Beeran et al. 2020; Nigam et al. 2020; Hosseini et al. 2021; Joshi and Prabhakar 2021; Patil and Nitave 2021; Warrier et al. 2021). The docking results for these herbal compounds were compared to those of the COX-2 inhibitor diclofenac. Among the compounds investigated, glycyrrhizin, rutin, solamargine, and solasonine demonstrated the most pronounced interactions with COX-2 (Table 2). Additionally, 3D docking images of the plant compounds interacting with the COX-2 protein are presented in Table S1. Among the compounds evaluated through docking studies, solasonine, solamargine, and rutin exhibited the most favorable docking scores of − 9.40, − 9.29, and − 9.10 kcal/mol, respectively. Subsequently, glycyrrhizin displayed a docking score of − 8.50 kcal/mol (Table 3). Solasonine demonstrated a total of ten hydrogen bonding interactions, comprising two hydrogen donors and eight hydrogen acceptors. The distances and energies of these interactions were measured in angstroms and kcal/mol, respectively. The amino acids involved in hydrogen donor interactions included PRo_514 (3.31 Å, − 0.7 kcal/mol) and GLN_192 (2.80 Å, − 0.6 kcal/mol). The hydrogen acceptor interactions involved SER_530 (2.61 Å, − 0.9 kcal/mol), ARG_513 (3.01 Å and 3.00 Å, − 0.8 and − 1.4 kcal/mol), ALA_516 (3.07 Å, − 0.5 kcal/mol), PRO_514 (3.54 Å, − 0.5 kcal/mol), HIS_95 (2.91 Å, − 2.7 kcal/mol), and GLY_354 (3.01 Å and 2.77 Å, − 0.5 and − 0.7 kcal/mol).

Table 1.

Phytochemical profiles of selected medicinal plants employed in the management of inflammation

No	Plant name	Herbal composition	Pubchem ID	References
1	Gmelina arbore	Luteolin	5,280,445	Warrier et al. (2021)
		Kaempferol	5,280,863
		Iso-quercetin	5,280,804
		Rutin	5,280,805
2	Coriandrum sativum	Linalool	6549	Hosseini et al. (2021)
		γ-terpinene	7461
		Neryl acetate	1,549,025
3	Glycyrrhiza glabra	Glycyrrhizin	114,982	Pastorino et al. (2018)
		Glabrin A	161,850
4	Terminalia chebula	Catechin	9064	Nigam et al. (2020)
		Gallic acid	370
		Quercetin	5,280,343
		Ellagic acid	5,281,855
5	Solanum nigrum	Solasonine	119,247	Gu et al. (2018)
		Solamargine	73,611
6	Vernonia cinerea	Vernolide A	5,281,508	Beeran et al. (2020)
7	Portulaca oleracea	Apigenin	5,280,443	Kumar et al. (2022)
		Malic acid	525
		Alpha-linolenic acid	5,280,934
8	Azadirachta indica	Azadirachtin	5,281,303	Joshi and Prabhakar (2021)
		Nimbin	108,058
9	Thespesia populnea	ß-Sitosterol	222,284	Patil and Nitave (2021)

Table 2.

Binding energies of compounds from nine medicinal plants targeting COX-2

No	PubChem CID	Compounds	Binding energy (KJ/mol)
1	119,247	Solasonine	− 9.40
2	73,611	Solamargine	− 9.29
3	5,280,805	Rutin	− 9.10
4	14,982	Glycyrrhizin	− 8.50
5	108,058	Nimbin	− 8.117
6	5,281,303	Azadirachtin	− 7.854
7	222,284	Beta-Sitosterol	− 7.844
8	5,280,804	Isoquercetin	− 7.678
9	5,280,934	Alpha-linolenic acid	− 7.522
10	9064	Catechin	− 6.909
11	70,690,655	Vernolide A	− 6.952
12	5,280,443	Apigenin	− 6.701
13	5,280,445	Luteolin	− 6.738
14	1,549,025	Neryl acetate	− 6.349
15	5,280,343	Quercetin	− 6.311
16	5,280,863	Kaempferol	− 6.219
17	5,281,855	Ellagic acid	− 5.784
18	3033	Diclofenac	− 5.685
19	6549	Linalool	− 5.591
20	161,850	Glabrin A	− 4.573
21	7461	γ-terpinene	− 4.089
22	370	Gallic acid	− 4.358
23	525	Malic acid	− 4.210

Table 3.

Insights into the docking interactions of optimal ligands within the active sites of COX-2

No	Compounds	H-bonding interactions and distance (Å)
1	Solasonine	PRo_514 (3.31)
		GLN_192 (2.80)
		SER_530 (2.61)
		ARG_513 (3.01)
		ALA_516 (3.07)
		PRO_514 (3.54)
		HIS_95 (2.91)
		GLY_354 (3.01)
		GLY_354 (2.77)
2	Solamargine	PRO_514 (3.41)
		HIS_95 (3.65)
		THR_94 (2.79)
		THR_94 (3.35)
3	Rutin	ARG_120 (3.02)
		ARG_120 (3.26)
		ALA_527 (3.29)
4	Glycyrrhizin	TYR_355 (2.69)
		ARG_513 (3.59)
		SER_353 (3.12)
		ALA_527 (2.96(
5	Diclofenac	TYR_385 (2.96(
		SER_530 (3.09(

Solamargine exhibited five hydrogen bonding interactions, consisting of one hydrogen donor and four hydrogen acceptors. The hydrogen donor was SER_353 (3.30 Å, − 0.5 kcal/mol), while the hydrogen acceptors included PRO_514 (3.41 Å, − 0.7 kcal/mol), HIS_95 (3.65 Å, − 0.6 kcal/mol), and THR_94 (2.79 Å and 3.35 Å, − 1.7 and − 0.5 kcal/mol). Rutin displayed three hydrogen acceptor interactions with residues, specifically with ARG_120 (3.02 Å and 3.26 Å, − 2.1 and − 0.9 kcal/mol) and ALA_527 (3.29 Å, − 0.5 kcal/mol). Glycyrrhizin demonstrated four hydrogen bonding interactions, comprising one hydrogen donor and three hydrogen acceptors. The hydrogen donor was TYR_355 (2.69 Å, − 0.5 kcal/mol), while the acceptors included ARG_513 (3.59 Å, − 0.7 kcal/mol), SER_353 (3.12 Å, − 0.9 kcal/mol), and ALA_527 (2.96 Å, − 0.5 kcal/mol). Diclofenac exhibited two hydrogen acceptor interactions with residues, specifically with TYR_385 (2.96 Å, − 1.3 kcal/mol) and SER_530 (3.09 Å, − 1.8 kcal/mol).

Notably, the best-performing compounds in their interactions with COX-2 shared several common residues. SER_530 was involved in both diclofenac and solasonine interactions, indicating its potential role in binding affinity. Additionally, SER_353 was present in both glycyrrhizin and solamargine, while ARG_513 was found in both glycyrrhizin and solasonine. Furthermore, HIS_95 and PRO_514 were common to solasonine and solamargine, highlighting the significance of these residues in enhancing binding interactions (Table 3). The stability of the docking scores for the examined compounds was notably superior to that of diclofenac. As indicated by their lower docking scores, solasonine, solamargine, rutin, and glycyrrhizin exhibit greater binding affinity for COX-2. Two-dimensional and three-dimensional images of the compounds are presented in Table S2.

Molecular dynamics studies

The binding affinities of solasonine, solamargine, rutin, and glycyrrhizin within the active site of the COX-2 receptor were examined using molecular docking techniques. It is important to acknowledge that molecular docking can produce results that may lack sufficient accuracy due to the omission of critical factors influencing ligand–protein interactions, such as conformational flexibility and solvation effects (Singh et al. 2022; Hoseinpoor et al. 2024a). To improve the reliability of the findings, advanced molecular dynamics simulations were employed, which incorporated essential parameters including solvent effects, ionic strength, pressure, and temperature variations (Rezvanpoor et al. 2023). Following a 100-ns molecular dynamics simulation, an assessment of the structural stability of the system was conducted through the calculation of key metrics, including root mean square deviation (RMSD), root mean square fluctuation (RMSF), and radius of gyration (Ahangarzadeh et al. 2022).

Root mean square deviation

The Root Mean Square Deviation (RMSD) serves as a fundamental metric for quantifying the average distance between atoms, which is crucial for assessing the structural stability and equilibration of the COX-2 protein in the presence of docked ligands. By comparing the RMSD values of the unbound (apo) protein to those of the ligand-docked complexes, researchers can elucidate alterations in the molecular dynamics of the protein and evaluate the conformational stability of the protein–ligand complexes (Hoseinpoor et al. 2024a; Ghahremanian et al. 2022). Figure 2 illustrates the evaluation of COX-2 protein stability under both ligand-bound and unbound conditions. The RMSD values for the carbon atoms were calculated and plotted over a simulation duration of 100 ns. The mean RMSD values obtained were as follows: for the apo-form, 0.216 ± 0.028 nm; for the complex with solasonine, 0.252 ± 0.025 nm; for solamargine, 0.231 ± 0.024 nm; for rutin, 0.241 ± 0.040 nm; for glycyrrhizin, 0.226 ± 0.025 nm; and for diclofenac, 0.289 ± 0.034 nm. Notably, both the ligand-bound complexes and the apo-form achieved a state of stability after approximately 30 ns of simulation, as depicted in Fig. 2. The binding of these ligands to COX-2 corroborates the presence of stable interactions. However, the relatively elevated RMSD observed in the ligand-bound complexes may be attributed to the increased conformational flexibility of COX-2 following ligand binding. Previous studies suggest that an RMSD range of 0.2–0.3 nm is indicative of stable protein–ligand interactions (Samandar et al. 2022). Based on the data presented in Fig. 2, the average RMSD of the COX-2 complex with the four ligands, including the approved drug, ranged between 0.22 nm and 0.28 nm, reflecting the stability of these complexes in the biological system.

Fig. 2.

RMSD plot comparing the binding stability of four selected compounds, solasonine (pink), solamargine (yellow), glycyrrhizin (dark blue), and rutin (red), relative to diclofenac (green, approved drug) and COX-2 alone (black)

Root mean square fluctuation

The Root Mean Square Fluctuation (RMSF) analysis is a pivotal method for assessing the dynamic behavior and flexibility of proteins, particularly in the context of ligand interactions (Shinde et al. 2014a). By quantifying the fluctuations of individual residues, RMSF provides valuable insights into the macromolecular flexibility of proteins and highlights potential regional structural modifications (Cob-Calan et al. 2019). As shown in Fig. 3, the RMSF plots for the COX-2 protein in both its apo-form and complexed states reveal significant variations in flexibility. The average RMSF values for the apo-form and complexes with solasonine, solamargine, rutin, and glycyrrhizin were measured at 0.127 ± 0.056 nm, 0.132 ± 0.057 nm, 0.120 ± 0.057 nm, 0.133 ± 0.059 nm, and 0.128 ± 0.049 nm, respectively. In comparison, the standard ligand, diclofenac, exhibited a higher average RMSF value of 0.151 ± 0.075 nm, indicating that the bioactive compounds tend to stabilize the protein structure more effectively than the standard inhibitor.

Fig. 3.

RMSF plot comparing the stability of four selected compounds, solasonine (pink), solamargine (yellow), glycyrrhizin (dark blue), and rutin (red), with diclofenac (green, approved drug) and COX-2 alone (black)

The observed reduction in RMSF values for the receptor residues within the complexes can be attributed to the strategic positioning of the ligands within the active site of COX-2 (Sarkar and Sen 2022). Key residues such as SER_530, SER_353, ARG_513, HIS_95, and PRO_514 are crucial for maintaining the stability of these interactions. The flexibility of the protein when bound to rutin and the standard ligand is nearly comparable, suggesting that both compounds engage effectively with the active site. However, the modestly lower RMSF values for the complexes of 4 compounds compared to that with diclofenac underscore the superior binding affinities of these compounds, which may enhance the rigidity of the protein structure and promote effective ligand binding (Hoseinpoor et al. 2024a). Furthermore, the interactions between COX-2 and compounds resulted in slight increases in RMSF values compared apo-form, particularly in the loop regions. Despite these increases, the overall protein structure remained intact, indicating that these ligands can form stable and appropriate interactions with the target (Sarkar and Sen 2022).

Radius of gyration

The Radius of Gyration (Rg) is a crucial metric for evaluating the structural equilibrium of proteins in both their bound and unbound states. It provides insights into the overall size of a molecular chain and reflects the compactness and flexibility of protein structures within a biological context (Parida et al. 2020). In this study, the Rg values for the apo-protein and various ligand complexes, specifically solasonine, solamargine, rutin, glycyrrhizin, and a standard ligand, were assessed. The measured Rg values were found to be 2.460 ± 0.011, 2.475 ± 0.008, 2.481 ± 0.009, 2.473 ± 0.008, 2.476 ± 0.011, and 2.491 ± 0.013 nm, respectively (Fig. 4).

Fig. 4.

Rg plot comparing the conformational properties of four selected compounds, solasonine (pink), solamargine (yellow), glycyrrhizin (dark blue), and rutin (red), with diclofenac (green, approved drug) and COX-2 alone (black)

The observed Rg values suggest a notable degree of structural compaction in the protein–ligand complexes. Such compaction is vital for understanding the interactions between these ligands and the COX-2 receptor, a key player in the modulation of inflammatory pathways (Hoseinpoor et al. 2024a). The elevated Rg values indicate that robust interactions between the ligands and the COX-2 receptor do not lead to significant conformational alterations or a decrease in compactness. Instead, these compounds demonstrate behavior similar to that of the standard ligand, suggesting a stable interaction profile that may enhance therapeutic efficacy. Moreover, the analysis of Rg during MD simulations enables the monitoring of structural changes over time. A higher Rg typically correlates with a more relaxed protein conformation, while a lower Rg indicates a more rigid and compact structure (Samandar et al. 2022). The findings from the MD simulations suggest that the binding of these ligands to the COX-2 target does not substantially affect its compactness, thereby preserving the receptor's structural integrity. This preservation is critical for the therapeutic potential of these compounds in treating inflammatory conditions, as it implies that they can enhance the receptor's functional capabilities without compromising structural fidelity.

Pred-hERG tool for cardiotoxicity prediction

hERG interaction prediction is essential for identifying cardiotoxicity risks. Pred-hERG software achieves 89–90% accuracy in assessing various compounds, offering results in binary and multi-class formats (Braga et al. 2015a; Parikh et al. 2023; Sharma et al. 2023a). In the present study, four compounds, solamargine, rutin, glycyrrhizin, and the approved drug diclofenac, were evaluated using the Pred-hERG server, with their structures represented in SMILES format. The results of these analyses are comprehensively presented in Table 4. The predicted pIC50 values reflect the expected potency of these compounds in inhibiting the hERG ion channel, a critical parameter in assessing potential cardiac toxicity during the drug development process (Shinde et al. 2014b; Sharma et al. 2023b). The binary prediction model evaluates the likelihood of a compound acting as a blocker of the hERG ion channel, based on a threshold value determined during the training of the model (Sharma et al. 2019; Tiwari 2023; Shelar et al. 2014; Dogra and Kumar 2015). Conversely, the multiclass prediction, also known as "Categorical Potency," estimates a compound's efficacy in blocking the hERG channel and categorizes it into specific potency classes. It is important to recognize that the confidence level indicated by the model represents its reliability in making predictions rather than the accuracy of those predictions. This distinction underscores the importance of considering confidence levels in predictive assessments whenever feasible (Iranshahy et al. 2017; Kumar et al. 2022; Dubey and Kashyap 2014; Modi and Soni 2023; Mohanasundaram and Antoneyraj 2025). The findings presented in Table 4 indicate that solamargine, rutin, and glycyrrhizin are classified as non-hERG blockers according to the binary model and exhibit weak blocking activity within the multiclass framework. In contrast, diclofenac is identified as a moderate blocker within the multiclass framework, demonstrating inferior efficacy compared to the three natural compounds in promoting optimal cardiovascular function.

Table 4.

Assessment of cardiotoxicity for the selected six compounds, along with diclofenac, utilizing the PredhERG server

Mol.: Molecule; ^*: Best compound

Accurate prediction of ADMET properties

The precise prediction of ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) parameters plays a pivotal role in the drug development process. By facilitating the early identification and removal of compounds with a high probability of failure, this approach significantly reduces development costs and accelerates timelines. Focusing on compounds with favorable ADMET profiles enhances the efficiency of the drug discovery pipeline, ultimately increasing the likelihood of clinical success (DiMasi et al. 2016; Shakour et al. 2023). In the current investigation, the ADMET characteristics of three selected compounds, glycyrrhizin, rutin, and solamargin, were evaluated alongside the approved drug diclofenac, despite existing literature on their properties (Malik et al. 2017; Ganga et al. 2019; Uzzaman et al. 2021; Srivastava et al. 2022). The assessment utilized the QikProp module from the Schrödinger Suite (Schrödinger, 2021) and the pkCSM tool (Pires et al. 2015). The toxicity analyses revealed that the tested compounds were non-sensitizing to skin, with toxicity levels in Tetrahymena pyriformis being comparable to those of diclofenac. This suggests a favorable safety profile for these compounds in preliminary screening (Table S3). Furthermore, hepatotoxicity assessments yielded promising results for both glycyrrhizin and rutin, indicating a reduced risk of liver-related adverse effects. The QikProp analysis further highlighted that rutin demonstrated ADME profiles aligned with those of 95% of approved pharmaceutical agents, underscoring its potential as a viable candidate for further development (Table S1 (a–e)).

However, the analysis revealed several violations in QikProp parameters for glycyrrhizin and solamargin. For instance, the molecular weight (mol_MW) of glycyrrhizin (822.94) and solamargin (854.04) exceeds the acceptable range of 130.0–725.0 for 95% of known drugs. Additionally, solamargin's total solvent-accessible surface area (SASA) of 1175.23 Å is also outside the recommended range of 300.0–1000.0 Å. The hydrophobic component of the SASA (FOSA) for solamargin is reported at 846.02 Å, surpassing the upper limit of 750.0 Å. Furthermore, its total solvent-accessible volume of 2394.606 Å exceeds the acceptable range of 500.0–2000.0 Å.

Other concerning parameters include the predicted polarizability (QPpolrz) of solamargin, which is 80.296 Å, exceeding the threshold of 70.0 Å, and the predicted hexadecane/gas partition coefficient (QPlogPC16) of 24.994, which is significantly above the maximum acceptable value of 18.0. The predicted qualitative human oral absorption scores for solamargin, rutin, and glycyrrhizin are − 0.917, 1, and 1, respectively, indicating poor absorption for solamargin. In contrast, the predicted percentages of human oral absorption are 11%, 0%, and 0%, respectively, which are significantly below the acceptable threshold of > 80% for high absorption. Lastly, the solvent-accessible surface area of fluorine atoms (SAfluorine) for solamargin is reported as − 0.736, which is also out of range (Christopher et al. 2021; Chikhale and Rishipathak 2025; Lolak et al. 2025).

These violations raise important considerations regarding the formulation and derivatization of these natural product leads. While some limitations may be addressed through chemical modifications or formulation strategies, others may pose inherent challenges that could affect the viability of these compounds in drug development. Therefore, a detailed discussion on how these factors can be mitigated is warranted in future studies.

MM-GBSA results

Binding affinities derived from docking calculations are frequently considered unreliable for ranking compounds (Sirous et al. 2019). To improve accuracy, it is recommended to incorporate solvation energy and surface accessibility area (Huang et al. 2006). This study employed MM-GBSA calculations to assess rutin in comparison to the approved drug diclofenac, as shown in Table 5. Both compounds displayed favorable binding interactions, with ΔG_binding values of − 11.316 kcal/mol for Rutin and − 35.190 kcal/mol for Diclofenac.

Table 5.

Binding energies of natural compound rutin and approved drug at the COX-2 active site via MM-GBSA

Entry	∆GBinda	∆GCoulb	∆GCovalc	∆GHbondd	∆GLipoe	∆GSolvGBf	∆GVdwg
Rutin	− 11.316	− 6.389	48.142	− 2.303	− 34.993	31.312	− 42.370
Diclofenac	− 35.190	12.528	6.805	− 1.250	− 24.309	14.478	− 40.768

a, Binding free energy; b, Coulomb energy; c, Covalent energy; d, Hydrogen binding; e, Lipophilic energy; f, the generalized born electrostatic solvation energy; g, Vad der Waals energy

In-silico CLC-Pred cell line cytotoxicity prediction results

The CLC-Pred tool is a widely recognized resource in the fields of cheminformatics and medicinal chemistry, specifically engineered to predict the cytotoxicity of compounds against various cell lines, along with their associated activity probabilities. This computational approach facilitates the identification of promising candidates for further pharmacological evaluation, thereby streamlining the drug discovery process (Riju et al. 2009). Predictions are quantified using Pa values, where values exceeding 0.5 indicate a significant likelihood of cytotoxic activity against the predicted cancer cell lines. In this study, derived compounds of three plants were analyzed, of which two were selected for cytotoxicity assessment across multiple cell lines using the CLC-Pred tool. The results were systematically organized in a table that included the respective cancer types, probabilities, and classifications of the cell lines (Table 6).

Table 6.

Predicting cytotoxicity of compounds in cancer cell lines with CLC-pred

Pa	Pi	Cell-line	Cell-line full name	Tissue	Tumor type
Glycyrrhizin
0.611	0.016	HL-60	Promyeloblast leukemia@	Haematopoietic and lymphoid tissue	Leukemia
0.550	0.029	NCI-H838	NSCLC#	Lung	Carcinoma
0.516	0.021	HepG2	Hepatoblastoma	Liver	Hepatoblastoma
0.523	0.057	A549	Lung carcinoma	Lung	Carcinoma
Rutin
0.567	0.004	Caco-2	Colon adenocarcinoma	Colon	Adenocarcinoma
0.536	0.020	SK-MEL-1	Metastatic melanoma	Skin	Melanoma
0.533	0.024	HL-60	Promyeloblast leukemia@	Haematopoietic and lymphoid tissue	Leukemia
0.528	0.036	NCI-H838	NSCLC#	Lung	Carcinoma
Diclofenac
.595	0.011	IGROV-1	Ovarian adenocarcinoma	Ovarium	Adenocarcinoma
0.556	0.014	OVCAR-8	Ovarian adenocarcinoma	Ovarium	Adenocarcinoma
0.550	0.015	OVCAR-4	Ovarian adenocarcinoma	Ovarium	Adenocarcinoma
0.550	0.015	HCC 2998	Colon adenocarcinoma	Colon	Adenocarcinoma
0.529	0.014	HOP-92	Non-small cell lung carcinoma	Lung	Carcinoma
0.531	0.017	786-0	Renal carcinoma	Kidney	Carcinoma
0.525	0.016	TK-10	Renal carcinoma	Kidney	Carcinoma
0.527	0.019	OVCAR-5	Ovarian adenocarcinoma	Ovarium	Adenocarcinoma
0.523	0.016	UO-31	Renal carcinoma	Kidney	Carcinoma
0.522	0.021	UACC-257	Melanoma	Skin	Melanoma
0.511	0.016	SR	Adult immunoblastic lymphoma	Haematopoietic and lymphoid tissue	Lymphoma

#: Non-small cell lung cancer, @: Promyelocytic Leukemia (APL)

The compound glycyrrhizin, obtained from Glycyrrhiza glabra Link, exhibited notable cytotoxicity against hepatoblastoma (Pa = 0.516, Hep G2) and lung carcinoma (Pa = 0.523, A549), both yielding substantial Pa values. Similarly, rutin, derived from Gmelina arborea Link, demonstrated strong activity against colon adenocarcinoma (Pa = 0.567, Caco-2) and metastatic melanoma (Pa = 0.536, SK-MEL-1). Both glycyrrhizin and rutin displayed high cytotoxicity (Pa > 0.5) against HL-60 (promyeloblast leukemia) and NCL-H818 (non-small cell lung cancer, NSCLC).

Importantly, these compounds showed no toxicity towards normal cell lines, including PrEC (prostate epithelial cells), and HFL1 (human fetal lung fibroblasts) for glycyrrhizin, as well as CRL-7065 (fibroblast) and HaCaT (keratinocyte) for rutin. Moreover, two compound showed no toxicity toward normal cell lined of HASMC (aortic smooth muscle) (Table 7). These findings suggest that glycyrrhizin and rutin could be viable candidates for cancer therapy; however, it is essential to regard cytotoxicity as a secondary observation and refrain from overstating therapeutic claims. Additional in vitro and in vivo studies are necessary to validate their efficacy and safety (Naeem et al. 2022).

Table 7.

Predicting cytotoxic activity of compounds on normal cell lines with CLC-pred

Pa	Pi	Cell-line	Cell-line full name	Tissue
Glycyrrhizin
0.134	0.098	PrEC	Prostate epithelial cell	Prostate
0.035	0.001	HFL1	Human foetal lung fibroblast	Lung
0.034	0.007	HASMC	Aortic smooth muscle	Muscle
Rutin
0.076	0.004	HASMC	Aortic smooth muscle	Muscle
0.103	0.033	CRL-7065	Fibroblast	Skin
0.074	0.061	HaCaT	Keratinocyte	Skin
Diclofenac
0.152	0.069	IMR-90	Embryonic lung fibroblast	Lung
0.086	0.046	HaCaT	Keratinocyte	Skin
0.025	0.013	HASMC	Aortic smooth muscle	Muscle

Discussion

This in silico study employed computational methods to analyze phytochemicals as selective COX-2 inhibitors, requiring no ethical approval or consent from human or animal subjects. It highlights the therapeutic potential of compounds such as glycyrrhizin, rutin, solamargine, and solasonine, which exhibit promising binding interactions with COX-2, a key enzyme in inflammatory processes.

Glycyrrhizin, derived from Glycyrrhiza glabra, has garnered attention for its multifaceted pharmacological properties, including anti-inflammatory, antiviral, and anticancer activities. The docking studies conducted in this research revealed that glycyrrhizin exhibited a binding affinity with a score of − 8.50 kcal/mol, indicative of its potential efficacy as a COX-2 inhibitor. This finding aligns with previous studies, such as those by Mohit Saini et al. (Saini and Malik 2023), which elucidated the molecular basis for glycyrrhizin's antiproliferative activity against COX-2. The presence of multiple hydrogen bonds between glycyrrhizin and critical active site residues of COX-2 further supports its potential as a natural anti-inflammatory agent. Moreover, glycyrrhizin's cytotoxic effects against various cancer cell lines, including lung carcinoma, have been documented, with studies indicating its comparable efficacy to standard chemotherapeutics like paclitaxel (Kumar Yadav et al. 2013). The non-toxic profile observed in HEK293T cell lines reinforces the safety of glycyrrhizin, making it a viable candidate for further therapeutic applications. Additionally, its documented activity against SARS-CoV-2 highlights its relevance in contemporary health challenges, suggesting that glycyrrhizin could serve dual roles in combating inflammation and viral infections (Zamzami 2023).

Rutin, a flavonoid glycoside found in Gmelina arborea and other plants, has been extensively studied for its antioxidant and antihypertensive properties. The current study demonstrated that rutin exhibited a favorable binding affinity to COX-2, with a docking score of − 9.10 kcal/mol, indicating its potential as an effective anti-inflammatory agent. Previous investigations have shown that rutin selectively induces cytotoxicity in cancer cells while preserving the viability of normal cells, making it particularly promising for cancer therapy (Pravin et al. 2024). Furthermore, the ability of rutin to inhibit key proteins involved in the SARS-CoV-2 life cycle, such as the main protease and RNA-dependent RNA polymerase, underscores its multifaceted therapeutic potential (Rahman et al. 2021). The ADMET properties of rutin, which indicate optimal solubility and a favorable safety profile, further enhance its candidacy for clinical applications (Ganga et al. 2019; Pravin et al. 2024; Rahman et al. 2021). The findings from this study reinforce the notion that rutin could be leveraged not only for its anti-inflammatory properties but also as a potential adjunct in cancer treatment.

Solanum nigrum, or black nightshade, is recognized for its rich array of secondary metabolites, including steroidal glycoalkaloids such as solasonine and solamargine. The in-silico analysis revealed that solasonine (− 9.40 kcal/mol) and solamargine (− 9.29 kcal/mol) may serve as potential COX-2 inhibitors. Previous studies have indicated that solasonine can inhibit interactions between critical proteins like p53 and mortalin, thus presenting its anticancer potential (Pham et al. 2019). Abdelatif Aouadi et al. reported that the antioxidant activity linked to phenolic compounds in Solanum nigrum enhances its therapeutic profile, suggesting a comprehensive approach to managing inflammation and oxidative stress. Their research indicates that the favorable pharmacokinetic properties, characterized by high gastrointestinal absorption and low permeability to the blood–brain barrier, suggest that compounds derived from Solanum may have minimal side effects, thereby supporting their potential for clinical application (Aouadi et al. 2024). In this study, the evaluation of pharmacokinetic properties, particularly ADMET, reveals that solamargine and solasonine exhibit limited effectiveness.

This study presents robust evidence supporting the potential of bioactive compounds derived from Gmelina arborea, Glycyrrhiza glabra, and Solanum nigrum as effective COX-2 inhibitors. Utilizing molecular docking and dynamics simulations, it was demonstrated that solasonine, solamargine, rutin, and glycyrrhizin possess superior binding affinities compared to the standard non-steroidal anti-inflammatory drug (NSAID) diclofenac. The stability of the ligand-receptor complexes, validated through molecular dynamics analyses, further highlights their therapeutic efficacy. These findings suggest that these compounds could serve as promising natural alternatives to conventional anti-inflammatory medications, with potential applications in the treatment of inflammatory conditions and cancer. Notably, this study offers several advantages, including the identification of safe and effective bioactive compounds, a reduction in the time and costs associated with drug discovery, and the potential for developing novel therapeutics with fewer side effects.

Conclusion

This in-silico study provides strong evidence supporting the potential of bioactive compounds from Gmelina arborea, Glycyrrhiza glabra, and Solanum nigrum as effective COX-2 inhibitors. Molecular docking and dynamics simulations revealed that solasonine, solamargine, rutin, and glycyrrhizin exhibit superior binding affinities compared to the standard NSAID diclofenac, with docking scores of − 9.40, − 9.29, − 9.10, and − 8.50 kcal/mol, respectively. The presence of multiple hydrogen bonds with critical active site residues further underscores the strength of these interactions, suggesting enhanced anti-inflammatory potential. Molecular dynamics simulations confirmed the stability of these ligand-receptor complexes. Analyses of RMSD, RMSF, and Rg demonstrated minimal structural fluctuations and stable ligand binding throughout the simulation period, reinforcing the reliability of the docking results. These findings highlight the structural integrity and compactness of the COX-2 receptor upon ligand interaction, which is crucial for maintaining therapeutic efficacy. The cardiotoxicity assessment using the Pred-hERG tool indicated a favorable safety profile for solamargine, rutin, and glycyrrhizin, classifying them as non-hERG blockers with a lower risk of cardiotoxicity compared to diclofenac. Additionally, ADMET predictions revealed promising pharmacokinetic properties, with rutin emerging as a particularly viable candidate due to its alignment with established pharmaceutical profiles. Furthermore, cytotoxicity predictions using the CLC-Pred tool suggested that rutin and glycyrrhizin exhibit significant activity against cancer cell lines while showing minimal toxicity toward normal cells. These findings suggest potential applications beyond anti-inflammatory activity, highlighting their promise as therapeutic agents for various diseases. Overall, rutin and glycyrrhizin stand out as promising COX-2 inhibitors with significant therapeutic potential for treating inflammatory conditions. Given their favorable pharmacokinetic and safety profiles, further in vitro and in vivo studies are essential to validate these computational findings and facilitate their development into clinically viable treatments.

Material and methods

Molecular docking studies

This investigation employed molecular docking techniques to elucidate the binding interactions between the COX-2 protein and twenty-two selected phytoconstituents of Gmelina arbore, Coriandrum sativum, Glycyrrhiza glabra, Terminalia chebula, Solanum nigrum, Vernonia cinerea, Portulaca oleracea, Azadirachta indica and Thespesia populnea. The three-dimensional structure of the human COX-2 receptor was acquired from the RCSB Protein Data Bank (PDB code: 1pxx) (www.rcsbPDB.org). The corresponding SDF files for the ligands were obtained from the PubChem database (http://pubchem.ncbi.nlm.nih.gov). The initial processing of ligand structures was executed using MOE2022 software (Hoseinpoor et al. 2024a; Ahangarzadeh et al. 2022). Structural preparation for the COX-2 receptor entailed modifications to the protein structure, the incorporation of polar hydrogen atoms, and energy minimization to establish an optimal environment for docking (Ahangarzadeh et al. 2022; Yadav et al. 2022). Water molecules were systematically eliminated to enhance the accuracy of binding site analysis. Docking was performed at the site of the cognate ligand. Docking simulations were performed utilizing the Triangle Matcher method for ligand placement, complemented by the London dG scoring function to calculate binding free energy. For each ligand, a total of 30 docking poses were generated, from which five optimal outputs were selected for further assessment. The resultant data, encompassing structural configurations and binding energies, were stored in mdb format (Rezvanpoor et al. 2023). The final evaluation of the optimal conformation was conducted based on the number of hydrogen bonds formed and the Gibbs free energy (binding energy) (Ahangarzadeh et al. 2022; Yousefi et al. 2025a). Subsequently, the MD modeling phase utilized the optimal docking states of the compounds alongside an approved drug for comprehensive analysis.

Molecular dynamics studies

Molecular dynamics (MD) simulations are an essential computational methodology for elucidating the dynamic behavior and conformational changes of proteins across a range of time scales (Rasouli et al. 2017; Yousefi et al. 2025b). In this study, MD simulations were executed using GROMACS 2018 software on a Linux platform, concentrating on the COX-2 receptor and its interactions with bioactive compounds. The preliminary phase involved the preparation of COX-2 protein files through the pdb2gmx tool, utilizing the CHARMM36 force field (Shakour et al. 2025). This process generated crucial files, including molecular topology, structural, and dynamic parameter files. Ligand topology files were subsequently created via the SwissParam web server and integrated with the protein files to construct a comprehensive topology file for the protein–ligand complex. To accurately simulate a physiological environment, the protein–ligand complex was positioned within a cubic box filled with SPC (Simple Point Charge) water molecules, ensuring a minimum separation of 1 nm between the protein surface and the box boundaries (Shakour et al. 2022). The system was neutralized by incorporating sodium ions and their counterions, chloride, to achieve charge balance, which is critical for accurate computational analysis. Energy minimization was performed using the Steepest Descent algorithm to resolve steric clashes and optimize the initial molecular configuration. The system was equilibrated for 200 ps using NVT (constant Number of particles, Volume, and Temperature) and NPT (constant Number of particles, Pressure, and Temperature) ensembles at 300 K and 1 bar (Hoseinpoor et al. 2024b). The simulation employed a 2fs time step, followed by 100 ns of MD simulation. Periodic boundary conditions were removed to facilitate trajectory analysis, allowing for the quantification of clusters formed during the simulation, with a total of 10,000 frames analyzed. The stability of the protein–ligand complex was rigorously evaluated through various analytical methods, including RMSD and RMSF analyses, which provided insights into the structural integrity and atomic fluctuations of the complex. Additionally, the Rg was assessed to determine the compactness of the protein structure. The complexes with the high degree of stability were selected for further investigation.

Pred-hERG tool for cardiotoxicity prediction

The hERG (human ether-a-go-go-related gene) encodes the alpha subunit of the potassium ion channel known as kV 11.1. This channel plays a crucial role in facilitating normal electrical signaling and ensuring optimal function of the human cardiovascular system. Inhibition of the hERG protein, which results in the blockade of potassium ion channels, can precipitate arrhythmias, a serious cardiac condition. Numerous compounds have been identified that can bind to the hERG protein, demonstrating the potential to induce cardiotoxicity and disrupt normal cardiac rhythm, potentially leading to fatal outcomes (Braga et al. 2014). Traditional assays for assessing hERG inhibition are often characterized by being time-intensive, costly, and labor-intensive, with significant limitations regarding throughput and scalability (Shakour et al. 2023; Hoseinpoor et al. 2024b). Consequently, it is imperative to predict and understand the cardiotoxic potential of pharmacological compounds prior to their approval for clinical use. High-throughput ion-channel screening data frequently exhibit considerable variability, which can affect the reliability and regulatory suitability of the results (Braga et al. 2015b). To address these challenges, the Pred-hERG web server tool (available at http://www.labmol.com.br/) has been developed to predict the cardiotoxicity of drug-like compounds. This tool is capable of distinguishing between hERG blockers and non-blockers, thereby providing valuable insights into the safety profiles of various compounds.

In this study, four specific phytoconstituents, solasonine, solamargine, rutin, and glycyrrhizin, were analyzed using the Pred-hERG server. These compounds were submitted in the form of SMILES (Simplified Molecular Input Line Entry System). The tool evaluates compounds for cardiotoxicity with an accuracy, sensitivity, and specificity of approximately 89–90% (Shakour et al. 2025). The results obtained from the Pred-hERG analysis were subsequently organized and presented in a tabular format, highlighting the potential risks associated with these bioactive compounds.

Accurate prediction of ADMET properties

The efficacy of a pharmaceutical compound is significantly influenced by its ADMET profile. Research indicates that approximately 40% of drug candidates fail in clinical trials due to inadequate ADMET properties (Kola and Landis 2004). These failures contribute to the rising costs associated with the development of new medications, underscoring the necessity of optimizing ADMET characteristics early in the drug discovery process (DiMasi et al. 2016). Empirical methodologies, particularly Quantitative Structure–Activity Relationship (QSAR) models, form the foundation for assessing ADMET properties. These models have proven effective across a wide range of compounds, establishing a strong correlation between molecular structure and biological activity using high-quality data (Shakour et al. 2021; Azzam 2023; Murali et al. 2023). In this study, the QikProp module from Schrodinger Maestro, which functions as the offline ADME server of the Schrodinger Suite, was utilized in conjunction with the pkCSM webserver to predict the ADME and toxicity profiles of three selected compounds, glycyrrhizin, rutin, and solamargine, along with the approved drug diclofenac (Shakour et al. 2025).This approach aims to enhance the understanding of the pharmacokinetic and toxicological properties of these compounds, facilitating more informed decision-making within the drug development pipeline (Raut et al. 2023; Thorat et al. 2023).

MM-GBSA study

The binding free energy of rutin and the standard compound was assessed using the MM-GBSA method within the Prime module of the Schrödinger Suite 2015 (Mulakala and Viswanadhan 2013). The optimal conformation of the ligand–protein complex was chosen for energy calculations, utilizing the OPLS-2005 force field alongside the generalized Born surface area (GBSA) method. The binding free energy for each ligand was calculated using the following equation (Abbasi et al. 2024):

In-silico prediction of cell line cytotoxicity with CLC-pred tool using PASS

Computer-aided structure–activity relationship (SAR) prediction studies are pivotal in drug design, particularly in the identification of novel biomarkers for disease treatment. The CLC-Pred Tools facilitate the prediction of cytotoxicity in tumor cell lines, leveraging structure-cell line cytotoxicity relationships derived from the Prediction of Activity Spectra for Substances (PASS) framework. This methodology utilizes specialized training sets and implements a leave-one-out cross-validation procedure to ensure robust predictive performance. PASS provides critical bioactivity assessments for chemical compounds, quantified as Pa (Probable Activity) and Pi (Probable Inactivity) values, which indicate the likelihood of a compound's activity status. Notably, the accuracy of in-silico predictions using this approach demonstrates a remarkable 96% agreement with in vivo experimental outcomes. The predicted cytotoxicity against various human cell lines is expressed through Pa values, where a Pa value greater than 0.5 signifies a high probability of activity, while Pi values denote inactivity (Riju et al. 2009).

Supplementary Information

Below is the link to the electronic supplementary material.

Authors contribution

F.A. & F.H. conducted some experiments; S. F. edited the main manuscript, R. AD. managed the project; N. SH. designed the project, wrote the main manuscript, prepared the figures, and conducted some experiments.

Data availability

The authors confirmed that the data supporting the study's conclusions are included in the article and its supplementary materials. Upon a reasonable request, the corresponding author will provide the raw data used to support the findings of this study.

Declarations

Conflict of interest

The authors declare no competing interests.