Literature reports on large molecule inhibitors, particularly with sulfonamide functional moieties, wherein the sulfonamide group is directly linked to 5- or 6-membered aromatic ring systems, are available (Pecina et al., 2018; Singh et al., 2025). However, non-sulfonamide inhibitors are rarely reported because a weaker interaction is exhibited with the binding pocket residues, along with significant toxicity values (Cheng et al., 2024; Pontecorvi et al., 2025). These non-sulfonamide inhibitors were found to possess chemical moieties like PO₃ ⁻, COS⁻, and COO⁻ (data not reported). Currently, investigations on sulfonamide-based inhibitors are being prioritized, exploring the possible gaps that can enhance the activity of the existing inhibitors. Several experimental reports offer outstanding inhibition potentials but lack mechanistic details. Herein, efforts were taken to investigate the introduction of linkers with different lengths between the sulfonamide moiety and aromatic rings. Moreover, literature reports have revealed that sulfonamide-based inhibitors exhibit stronger interactions and better inhibition potential than non-sulfonamide derivatives. Hence, this study might emphasize the key atomistic interactions that can influence the inhibition potential of sulfonamide derivatives effectively.

To narrow the library of inhibitor ligand molecules, pharmacophoric analysis was carried out on a relative scale through the assignment of positives and negatives.

The pharmacophore hypothesis for various CAII inhibitors was constructed using the phase module (Schrödinger suite 2023–2024). From the values of dock score and binding free energies, thresholds were assigned as the highest and lowest affinity scores. Non-covalent interactions with receptor sites could be simplified by positioning the pharmacophoric regions within a defined three-dimensional space. The pharmacophoric sites were constructed by utilizing the hydrogen bond donor (D), hydrogen bond acceptor (A), hydrophobic group (H), and positive (P) and negative (N) ionizable and aromatic ring (R), as described in the module phase. Of 186 ligands, 56 (32 sulfonamides and 24 non-sulfonamides) were selected as training sets by their wide range of activity and diversity, as predicted by the observed best and poor docking scores. The other 130 compounds served as test sets to facilitate a meaningful comparison with the predicted models. The pharmacophore hypotheses were assessed using their scoring function to maximize the alignment of ligands with high affinity while also including characteristics from ligands with low affinity to enhance the model’s versatility. The pharmacophore hypothesis was produced using 29 active compounds, with 1Å denoting the pharmacophore-matching tolerances among two pharmacophoric characteristics and 2Å representing the lowest inter-site distance. Ten different hypotheses were developed, with two serving as the maximum number of sites, with 50% serving as actives. The best of the ten produced hypotheses was chosen based on the number of matches, volume, vector, energy terms, site score, survival inactive, and survival score (Table 1). Different pharmacophore models (A2, A3, D6, R8, and R9) are shown in Figure 2, and the hypotheses scores are given in Table 1. The best survival score (AADRR_1) indicates a better alignment of the active ligand with the target site and will be a potent pharmacophore model. In the generated pharmacophore hypothesis, the ligands with higher survival scores are known to exhibit closer geometric alignment with the model’s essential features, indicating their fit to the binding template is with minimal distortion. Similarly, a higher vector score signifies that directional interactions like hydrogen bonds or metal coordinations are oriented optimally toward the receptor site. Together, these metrics will help to identify compounds whose chemical features are correctly positioned and pointed to engage the protein efficiently. The alignment of various sulfonamide inhibitors along with the various pharmacophoric features is depicted in Figure 3 (also in Supplementary Figure S1) and associated scores for the best hypothesis are given in Table 2 (see Supplementary Table S3 for the parametric scores of AADRR_3).

Similarly, different pharmacophore models and their correspondingly derived hypotheses are shown in Figure 4, and associated features are given in Table 3. The alignments of various non-sulfonamide inhibitors along with the various pharmacophoric scores are depicted in Figure 4 (also in Supplementary Figure S2) and Table 3. The associated scores for the best hypothesis (AADR_1) are given in Table 4 (also see Supplementary Table S4 for the parametric scores of AADH_1). The alignments of various non-sulfonamide inhibitors along with the various pharmacophoric features are depicted in Figure 5.

Further investigations were carried out with the identified ligand structures, focusing on the functional group fragment. The three-dimensional structures of ligands were generated using Open Babel software (Pandey and Pandey, 2024), with further modifications considering recent synthetic advances in heteroaryl functionalization [Ref]. Organic Letters, 27(22), 5625-5631. 10.1021/acs.orglett.5c01350. Herein, the influence of additional linking moieties like CH₂-, -NH-, -CH₂-CH₂-, and -NH-CH₂- on the binding interactions with the active site of CAII was investigated. As a proof of concept, an initial rigid potential energy surface (PES) scan was performed for the interaction of representative ligands with the residues localized within 5 Å from Zn²⁺ of CAII by varying the bond angle and dihedral angle using the B3LYP/6-31G method (Figure 6). The observation of a barrier along the surface indicates that the structural features of the ligands significantly influence their interaction with the active site residues, along with Zn²⁺. Thus, the observed variations in energy along the potential energy surface revealed that tuning the nature of functional groups on the ligands could pave the way to discover novel inhibitors. The induction of steric hindrance on the HN-S-NH moiety due to a change in the dihedral angle is shown in Figure 7.

In Figure 6A, the observed barrier (∼3.9 eV) could be due to the increase in the steric factor as the bond angle increases. The binding cavity void may not be large enough for any significant fluctuations of the interacting functional groups. The same is true for the variation in dihedral angle (Figure 6B). In the latter case, it was observed that at particular orientations of the functional groups with respect the aromatic skeleton, the strain energy has increased, which could be the possible reason for the observed barriers of 655.60 eV, 642.36 eV, 655.60 eV, and 559.09 eV, respectively, for the -NH-SO₂-NH-, -NH-CH₂-SO₂-NH-, -CH₂-SO₂-NH-, and -CH₂-CH₂-SO₂-NH- moieties along the PES. The reported binding pose of the benzenesulfonamide prong in the active site of human carbonic anhydrases I and II is shown in Figure 8. It can be seen that the amine group of the benzenesulfonamide ligand is coordinated to the active site Zn²⁺ (Elsayad et al., 2025).

Screening of the generated library of small-molecule ligands (sulfonamide and non-sulfonamide) identified from the ZINC database was carried out to obtain deeper insight into the binding interactions with the active site of CAII via molecular docking investigations (Irwin et al., 2012). For this purpose, the GLIDE module of Schrödinger was employed. From the library, 186 ligands were shortlisted and ranked by their dock score (Supplementary Table S1; Supplementary Figure S3), and the top 10 ligands were considered for further investigations (Figure 9). Based on the values of dock score, the effective binding of the above ligands with CAII was ranked, and their binding modes were evaluated. The structural details of all the ligands are given in Supplementary Table S2. The pertinent results are given in Table 5. The binding site was mostly composed of polar residues such as Lys131, Gln135, and Thr198, non-polar amino acids like Gly5, Pro199, and Val209, and aromatic residues such as Phe92, His121, and Phe94. The list of binding site interacting residues is given in Table 6. To narrow the investigation, the top three ligands in each of the sulfonamide and non-sulfonamide derivatives were chosen for evaluating their dynamic interactions with CAII. A strong π–cation interaction was observed for all the selected ligands that were actively coordinated to the Zn²⁺ of CAII. In the case of non-sulfonamide ligands, interactions were observed with residues including His93, Thr197, and Thr198 in addition to the observed π–alkyl interaction with Val120, Leu139, Val141, and Leu196. The sulfonamide ligands also exhibited strong hydrogen bonding interactions with Asn61 and Asn66 and π–sulfur interaction with His93. Moreover, the residues such as Val120, Val141, and Leu196 formed a π–alkyl interaction with the phenyl ring of the ligands. The best docked poses of sulfonamide (S6, S7, and S8) and non-sulfonamide (S1, S2, S15, and S16) ligands are shown in Figure 9, while the remaining ligands are shown in Supplementary Figure S1.

The scores from various free energy parameters that contribute significantly to the overall binding free energy of the docked interactions are given in Table 5. For example, the van der Waals contribution captures the close contact and steric interactions between the ligand and receptor. A good packing contribution is recognized with a significantly negative score. Favorable hydrophobic interactions depict the exposure of the ligands inside buried pockets. Good packing of hydrophobic groups in hydrophobic environments is energetically favored. These interactions reduce solvent exposure for non-polar atoms. This also mimics the free energy gain from hydrophobic packing.

The various residues of CAII that were effectively involved in the binding interactions with the ligands are given in Table 6. The binding pocket is mainly composed of polar residues like Asp, Thr, and Gln and non-polar residues, including Val, Leu, and Ala. Based on the dock score, sulfonamide ligands S6, S7, and S8 and non-sulfonamide ligands S1, S2, S15, and S16 were ranked sequentially. Among all the shortlisted ligands, S6 was observed to be exceptional with regard to its interaction with Zn²⁺. In this case, a strong π–cation interaction was observed with the aromatic ring, in addition to multiple strong hydrogen bonding interactions with Asn61, Asn66, Gln91, and Thr198. The benzimidazole moiety of the ligand was involved in π–π interaction with His93, while residues Val120, Val141, and Leu196 were involved in π–alkyl interactions. With all other ligands (S7 and S8), coordinate bonding interactions were observed with the sulfonamide group of the ligands and the Zn²⁺ of CAII (Figure 9). In S7, the interactions were contributed through four strong conventional hydrogen bonds with Gln91, His93, Thr197, and Thr198. In S8, Asn66, Gln91, and Thr198 formed three strong hydrogen bonds with the sulfonamide moiety, while interactions were observed with the Zn²⁺ and the sulfonamide group in addition to the π cloud of the ligand. Other interactions, like π–sulfur and π–alkyl interactions, were also involved in the stabilization.

In the case of non-sulfonamide ligands, Zn²⁺ was strongly coordinated to the –SO₂-NH- moiety (Figure 9). In all the cases, strong hydrogen bonds were observed with residues His93, Thr197, and Thr198, in addition to other interactions like π–π and π–alkyl interactions. The free energy parameters associated with the best docking interactions are given in Table 5 (Supplementary Table S1).

Although the docking investigations have revealed the effectiveness of binding interactions, the influence of dynamic motions under physiological conditions was not taken into account during the evaluation. Therefore, further investigations into the stability of interactions were carried out with the help of molecular dynamics simulations. These interactions are not static, and it is necessary to evaluate the variations in free energy and stability through dynamic simulations. For the purpose of understanding the dynamic behavior of the bio-molecular systems at their atomistic levels, molecular dynamics simulations were performed for the selected protein–ligand complexes. A detailed analysis of the key dynamic parameters, including the root mean square deviation (RMSD), radius of gyration (Rg), root mean square fluctuation (RMSF), hydrogen bonding, and solvent-accessible surface area (SASA), was carried out for the CAII protein and its interactions with sulfonamide and non-sulfonamide ligands. Here, the stability of the complexes under dynamic environments was evaluated for a period of 100 ns.

The root mean square deviation (RMSD) was evaluated as a measure of the influence of various ligands on the structural integrity of CAII and to quantify the changes incurred in the protein backbone throughout the simulation period. No significant changes were observed in the RMSD of the protein upon the approach of various best docked sulfonamide and non-sulfonamide ligands. The results revealed that the system under investigation is more stable during the interaction period. RMSD evaluations for the protein in the presence and absence of ligands were compared while assessing the stability of the complexes. The RMSD trajectories for various protein–ligand complexes are depicted in Figure 10. The average RMSD values for the complexes of S6, S7, and S8 (sulfonamides) were 1.26 ± 0.38 Å, 1.38 ± 0.37 Å, and 1.23 ± 0.46 Å, and the average RMSD values for S1, S2, S15, and S16 (non-sulfonamides) were 1.53 ± 0.56 Å, 1.39 ± 0.55 Å, 1.51 ± 0.69 Å, and 1.38 ± 0.69 Å respectively. The fluctuations observed in the trajectory during the initial simulation period were ignored as they may contribute to the stabilization phase, where the protein–ligand complex adjusts from its starting structure toward a thermodynamically favorable conformation under the field of applied force. These early phases include the relaxation of steric clashes, redistribution of solvent molecules, and adjustment of functional group orientations for meaningful structural deviations of the equilibrated complex (Grossfield et al., 2018; Schreiner et al., 2012; Walton and Vanvliet, 2006). Detailed analyses of the trajectories have shown that several factors and forces influence the stability of interactions, including solvent bridges, hydrophobic, π–alkyl, and hydrogen bonding. The extents of various interactions that contribute to the stability are depicted in Figure 11.

The various types of stabilizing interactions that influenced the binding of different ligands to the binding cavity, along with their percentage contributions, are depicted in Figure 11. In the case of S6, major stabilization was observed from the hydrophobic and π–π interactions involving residues like Val120, Val141, Leu196, Trp207 (hydrophobic), and His93 and His118 (π–π), although hydrogen bonding and water bridges also showed significant stabilizations. For S7, Asn61, Arg57, His93, and Thr198 revealed maximum stabilizing contributions through hydrogen bonding, hydrophobic forces, ionic forces, π–π forces, and solvent bridges, respectively. In S8, residues including Thr197, Leu196, Lys38, Lys223, and His93 were involved in hydrogen bonding, hydrophobic forces, ionic forces, and π–cation forces, and solvent bridge interactions, respectively.

In the case of non-sulfonamide ligands, except S1, other ligands revealed significant contributions through π–cation interactions. Residues like His93, Glu105, Leu196, and Thr198 showed significant contributions through π–π forces, hydrogen bonding, hydrophobic forces, and solvent bridges, respectively. The % contributions with other ligands are depicted in Figure 11.

RMS fluctuations are another parameter to evaluate the dynamic stability of the complexes. In this study, the positional fluctuations of all amino acid residues of the selected protein–ligand complexes were observed and are depicted in Figure 12. There was no notable fluctuation observed for the selected complexes except for S2 and S7. His2, His3, Lys8, His9, Asn10, Gly233, Gln234, Pro235, and Lys258 showed fluctuations >2 Å, and most of the residues that interacted with the ligand were located inside or close to the binding pocket. While analyzing the trajectory of the protein in the presence and absence of any ligand, only minor fluctuations were observed across all residues, with an average RMSF of 0.681 Å ± 2.22 Å. The average RMS fluctuations was observed for complexes S1, S2, S3, S6, S7, S8, S11, S12, S13, S14, S15, and S16 were 0.78 ± 1.24 Å, 0.743 ± 1.92 Å, 0.86 ± 2.34 Å, 0.67 ± 2.72 Å, 0.71 ± 1.08 Å, 0.71 ± 2.49 Å, 0.70 ± 1.14 Å, 0.68 ± 2.72 Å, 0.70 ± 2.13 Å, 0.67 ± 2.48 Å, 0.73 ± 2.61 Å, and 0.73 ± 2.25 Å, respectively.

The influence of various ligands on the stability of the complexes was evaluated by analyzing the radius of gyration during the 100 ns contact time. This helps to assess the protein’s structural compactness during interaction with different ligands. Rg is used to calculate the distance between the ligand’s central axis and the atom’s rotational position at which the most energy was transferred. It is well known that the Rg values are influenced by the conformational changes that are brought about by the binding of various ligands to the target. Here, the values of Rg for all the selected protein–ligand complexes were examined. Figure 13 revealed the trajectory of Rg during the evaluation period. The average radius of gyration was found to be 2.78 ± 0.11 Å, 2.86 ± 0.16 Å, 2.83 ± 0.16 Å, 3.37 ± 0.24 Å, 3.31 ± 0.61 Å, 2.97 ± 0.57 Å, 3.12 ± 0.37 Å, 2.82 ± 0.16 Å, 3.01 ± 0.26 Å, 2.45 ± 0.10 Å, 3.36 ± 0.13 Å, and 3.16 ± 0.30 Å, respectively, for the S1, S2, S3, S6, S7, S8, S11, S12, S13, S14, S15, and S16 complexes.

The interaction of ligands with CAII has a significant impact on the characteristics of the polar and non-polar surfaces of the ligand. The SASA analysis helps to locate the crucial binding pocket residues on the protein target and evaluate their accessibility to solvent molecules. Figure 14 shows the SASA trajectory. Supporting the earlier observations, the SASA trajectories also showed that non-sulfonamide ligands interact more strongly than sulfonamide derivatives, although significant fluctuations were detected at multiple points over a 100 ns simulation period. The position of the ligands at various time points during the course of simulation is also depicted in Figure 14. The observed fluctuations could be due to the presence of freely rotatable phenyl moieties on non-sulfonamide ligands rather than sulfonamide ligands. The average SASA for S1, S2, S3, S6, S7, S8, S11, S12, S13, S14, S15, and S16 complexes was found to be 42.91 ± 33.14 Å, 111.87 ± 269.44 Å, 37.53 ± 55.66 Å, 20.21 ± 51.16 Å, 159.48 ± 294.18 Å, 242.83 ± 177.80 Å, 125.66 ± 310.71 Å, 51.95 ± 71.81 Å, 213.89 ± 307.96 Å, 71.39 ± 274.949 Å, 333.46 ± 218.05 Å, and 354.87 ± 201.508 Å, respectively. The data from SASA assessments suggested that the solvent significantly contributes to stabilizing ligands as they bind to the cavity.

A variety of interactions preserve the structural integrity and stability of biomolecules. Among all the interactions, hydrogen bonding plays a major role in stabilizing the protein–ligand complexes. Here, investigations on the hydrogen bonding interactions were carried out for 100 ns, and the results are depicted in Figure 15. The factors responsible for stability are listed in the RMSD results. Among them, the hydrogen bonding contribution is greater because comparatively more energies are involved in binding. The results showed that the lifetime of the hydrogen bond indicates its strength. Several weak interactions, like hydrogen bonding and van der Waals interactions, are involved in energy transfer and enhance stability. The ability of various ligands to interact with the binding pocket residues of CAII is represented by their potential surface, as shown in Figure 16.

The above evaluated ligands were further investigated for their druggability through their pertinent pharmacokinetic characteristics, including absorption, distribution, metabolism, and excretion. The QikProp module of Schrödinger was utilized for the purpose. As a preliminary evaluation, Lipinski’s rule of five was taken into account along with star, central nervous system, human oral absorption, solvent accessibility, etc.

The star value represents the degree to which the ADME characteristics of a drug differ from those of 95% of other known inhibitors. The number of non-trivial, non-hindered rotatable bonds is represented as rotors (6.0–15.0). The CNS score ranges from −2 to +2, which influences ADME processes by regulating drug absorption through the blood–brain barrier, distribution via selective transport mechanisms, metabolism by brain-specific enzymes, and excretion through cerebrospinal fluid and efflux transporters. Molecules with a human oral absorption (HOA) value exceeding two are considered strong candidates for drug development. Lipinski’s rule of five plays a crucial role in ADME by providing guidelines to predict a compound’s oral bioavailability based on its molecular properties, including molecular weight (130.0–725.0), number of hydrogen bond donors (0.0–6.0), number of hydrogen bond acceptors (2.0–20.0), octanol water coefficient (log P_o/w = −2.0–6.5), and solubility (log S = −6.5 to 0.5), which collectively influence absorption and distribution characteristics. In ADME, solvent-accessible surface area (SASA) falls within the range of 300–1000, which is crucial as it influences a drug’s solubility, permeability, and interaction with biological membranes and proteins, thereby affecting its absorption and distribution properties. Polrz refers to the distribution of electric charge within a molecule and can influence a drug’s solubility, permeability through biological membranes, and interactions with proteins, which are critical factors in its ADME properties (polrz = 13.0–70.0). Hexadecane/gas coefficient is represented by log P_C16 (4.0–18.0). logHERG predicted IC₅₀ value represents the concentration at which a substance inhibits HERG K⁺ channels (log HERG = below −5). PCaco helps predict a compound’s permeability through the intestinal epithelium, indicating its potential oral bioavailability and absorption by assessing the five Lipinski’s rule properties (<25 poor, >500 great). Log BB value denotes the predicted brain/blood partition coefficient (−3.0 to 1.2). From the above evaluations, the compounds S1, S2, S3, S7, S11, and S14 did not show favorable pharmacokinetic properties.

The skin permeability coefficient is determined in terms of logK_p (−8 to −1). The water–gas partition coefficient should lie between 4 and 45. The drug binding with human serum albumin was determined in terms of logK_HSA, and it ranges between −1.5 and 1.5 (Colmenarejo, 2003). Apart from the above-mentioned exceptions, all the other compounds have revealed their drug potential. The above parameter values are tabulated in Table 7. The ZINC database compounds that satisfy all criteria were utilized for subsequent stages of docking and MM-GBSA studies. The relevant data pertaining to the toxicity of the reported derivatives, along with a positive control acetazolamide, are compiled in Table 8 (Supplementary Table S5) (ref: International Immunopharmacology, 134, 112,178. https://doi.org/10.1016/j.intimp.2024.112178).

Based on the values of star, molecular weight, hydrogen bonding capacity, and number of rotational bonds, the drug likeness is predicted as all the molecules to possess good drug-likeness (Star = 0), except S14. The hydrogen bond donor/acceptor values are within standard ranges, with S16 having the highest count with 3 donors and 6.7 acceptors. From the values of log Po/w, S1 is observed to be extremely lipophilic (9.793), which will have a significant influence on the solubility and permeability, while S14 is more hydrophobic (−1.45). All the molecules except S1 (1.786) have moderate aqueous solubility. All the molecules have a low risk of cardiotoxicity, although S7 and S8 are high. The intestinal permeability is also in range for all the molecules, while S8 and S15 reveal excellent permeability. S8 and S15 have good oral absorbability, while S6 and S12 show a lower probability of absorption. S2, S3, S13, and S14 display potent CNS activity. S14 and S13 have potent blood–brain barrier (BBB) penetration, while others do not effectively cross the BBB. Most of the compounds show low dermal permeability, with S8 being the least permeable and S1 and S2 being the most permeable. Based on the above observations, it can be concluded that S8 (sulfonamide) and S15 (non-sulfonamide) are the best overall candidates with high oral absorption, good permeability, and moderate toxicity risk. S14 and S13 have favorable BBB penetration and CNS scores. Compounds S6, S11, and S12 have moderate PHOA and low permeability. Because all the compounds present unique profiles, they may suit different therapeutic needs. Further in vitro and in vivo validations are recommended to confirm these predictions.

ProTox-II, an online browser-based toxicology prediction program, was utilized to predict the toxicity targets for the different ligands (Banerjee et al., 2018). Acute toxicity, toxicity endpoints, organ toxicity, stress response pathways, nuclear receptor signaling pathways, and other toxicity endpoints were all predicted. For validation, this tool uses selective oversampling in conjunction with the probability-based CLUSTER cross-validation. This tool was used to assess the possible risks, namely, the immunotoxicity, mutagenicity, cytotoxicity, carcinogenicity, and hepatotoxicity, associated with different sulfonamide and non-sulfonamide ligands. The predicted toxicity parameters for the shortlisted ligands are given in Table 8.

The results indicate that most compounds exhibit moderate toxicity profiles, with predictions varying from high acute toxicity for S6 to low acute toxicity for S14. Correspondingly, S6 and S7 fall into class-2 toxicity (more toxic), while S14 falls into class 6 (least toxicity). Compounds S12, S14, and S16 show relatively low hepatotoxicity and carcinogenicity probabilities, whereas S2 and S3 have higher values (≥0.80), suggesting a greater risk for liver damage and cancer potential. High immunotoxicity and mutagenicity scores are observed for most compounds (especially S1–S3), with S6 being comparatively lower in mutagenicity.

Receptor-based toxicity predictions indicate that nearly all compounds interact strongly with nuclear receptors, particularly AR, ER, PPARγ, and AhR, suggesting potential endocrine or metabolic pathway interactions. Notably, compounds also display significant activation of stress response elements, such as Nrf2/ARE and HSE, which may imply oxidative or heat shock stress responses. Mitochondrial toxicity, indicated by mitochondrial membrane potential (MMP) values, is highest for S11 and S12. Finally, most compounds show strong interactions with tumor suppressor p53 and DNA damage-associated protein ATAD5, hinting at possible genotoxic stress.

Structurally, the higher toxicity of S6 and S7 could be attributed to their heteroaromatic-rich sulfonamide scaffolds. S6 bears a benzimidazole–sulfonamide framework that furnishes an extended, planar π-surface and basic nitrogens, promoting off-target π–cation–π contacts and cationic amphiphilicity, consistent with its class-2, very low LD₅₀ value. Meanwhile, S7 exhibits the highest H-bond acceptor load (6) and multiple donors (3), which can increase interaction promiscuity and align with the increased risk for hERG as observed in silico; together these features rationalize its class-2 acute toxicity assignments. In contrast, S8 maintains the sulfonamide Zn-binding pharmacophore without these liabilities and yields class-4 toxicity and a more favorable ADME/toxicity profile. Therefore, S6 and S7 present the highest toxicity concerns due to their low LD₅₀ and toxicity class, while S16 appear the safest based on acute toxicity, hepatotoxicity, and lower receptor activation profiles.