QSAR study
A combined GFA and MLR approach led to the selection of four (4) descriptors in the generation of the QSAR model. The built model (Eq. 14) was found to excellently satisfy the requirement for a reliable QSAR model. The low residual values between the experimental and predicted activities as shown in Table 6 indicate a high predictive strength for the QSAR model. The R2 values of 0.801 and 0.892 for training set and test set, respectively, as obtained from the plot of exp. pIC50 against pred. pIC50 in Fig. 1 compare perfectly well with those obtained from GFA (0.8013 and 0.892) and MLRplusValidation analysis (0.8013 and 0.892) as reported in Table 5. The close grouping of points along the line of best fit in Fig. 1 shows a very good correlation between the experimental and predicted activity values, indicating that the built model is reliable and robust. The random spread of standardized residuals on both sides of the zero mark in Fig. 2 is an indication that the built model is free of any systematic error.
Furthermore, the low correlation coefficients (less than 0.50) which exist between each pair of the descriptor in the built model (Table 7) indicate no inter-correlation between each descriptor. Similar result was also obtained elsewhere by Adeniji et al. (2018) and Abdullahi et al. (2022). The VIF values ranging from 1 to 5 for all 4 descriptors as reported in Table 7 showed that the descriptors were statistical orthogonal and the built model was statistically substantial, an indication of the stability and acceptability of the built QSAR model. Similar observation was reported by Adeniji et al. (2019). The values of the absolute t-statistics greater than 2 for each descriptor show that the selected descriptors were good (Adeniji et al. 2018). Also, the evaluated p-values for the various descriptors in the model at 95% confidence level were less than 0.05 as shown in Table 7. Therefore, the alternative hypothesis which asserts that a relationship exists between the descriptors used in generating the model and the compounds’ inhibitory activities at p ˂ 0.05 holds. Additionally, the values of the mean effect (ME) reported in Table 7 provide vital information on the effect and degree of each descriptor’s contributions in the built model. The magnitudes and signs of ME values signify their individual strength and direction on the molecules’ inhibitory activities. All the descriptors except ATSC6v have positive ME, meaning that increase or decrease in their values will lead to an increase or decrease in the anti-proliferative activities, respectively. Increasing the values of ATSC6v will lead to a decrease in the inhibitory activities because of its negative ME value. GATS5c with the highest ME value has the greatest influence on the molecules’ inhibitory activities. GATS5c is Geary autocorrelation of lag 5 weighted by gasteiger charge, which has a positive ME suggested to contribute positively to anti-leishmanial activity. The gasteiger charge is a physicochemical property calculated for every atom in the molecule and is the Geary coefficient (Mahmud et al. 2020).
The low values of R2 and Q2 obtained from the random reshuffling (Table 8) inferred that the built model is stable, robust and reliable. The value of coefficient for Y-randomization, cR2p (0.664372) greater than 0.50, supports the claim that the built model is powerful and not inferred by chance. The William’s plot (Fig. 3) clearly shows that all the compounds fall within the square area ± 2.5 of standardized cross-validated residual. It can therefore be inferred that no outlier is present in the dataset. However, compound 1 was found with leverage value greater than the calculated warning leverage (h* = 0.75) and was said to be an influential molecule.
Virtual docking screening
Binding energies of the protein–ligand (drug) interactions are important to describe how well the drug binds to the target macromolecule. The negative value of the binding energy change shows the spontaneity of the binding process and how well ligands can fit into the target protein pocket to form the most energetically stable drug receptor (Ugbe et al. 2021). Among the studied receptors, pyridoxal kinase (PdxK) receptor (pdb id: 6K91) had relatively shown the strongest interaction with the various compounds as shown by the higher binding energy values associated with this receptor (Table 9). Consequently, PdxK was selected as the target receptor of interest in this study and subsequent discussions pertaining protein–ligand interactions shall be based on it. Also, among the 28 maleimides studied, compounds 14, 21 and 24 bind more strongly to PdxK with the highest reported binding energies of − 7.7, − 7.7 and − 7.8 kcal/mol, respectively. These compounds were equally well predicted by the built QSAR model with low residual values and contained within the model’s applicability domain (Table 6 and Fig. 3). Therefore, 14, 21 and 24 were selected as template molecules for designing new compounds with improved binding scores and pharmacological properties.
As seen from Table 10 and Figs. 5, 6, 7, 8, compound 14 was observed to have interacted well with the binding site of the PdxK receptor through three (3) conventional hydrogen bonds, one (1) carbon-hydrogen bond, one (1) π-anion, one (1) π–π stacked and one (1) pi-alkyl interactions. One of the carbonyl oxygen atoms on the pyrrole ring system formed 3 conventional hydrogen bonds with 2 THR-229 at distances of 2.41 Å and 2.48 Å, and GLY-228 at a distance of 2.37 Å. It also formed a carbon-hydrogen bond with ASP-231 at a distance of 3.44 Å. Others include π-anion between the pyrrole ring system and ASP-124, pi-pi stacking between the phenyl ring and aromatic ring of TYR-85, and π-alky interaction between the phenyl ring and VAL-19. Similarly, Compound 21 binds well into the binding pockets of the PdxK receptor via five (5) conventional hydrogen bonds, a carbon-hydrogen bond, one π-anion, two π-alkyl, and one alky interactions. One of the carbonyl groups oxygen formed conventional hydrogen bonds with GLY-228 and 2 THR-229 at a distance of 2.6 Å, 2.15 Å and 2.47 Å, respectively. The other carbonyl oxygen interacted with ASN-151 and LYS-187 to form conventional hydrogen bond at a distance of 2.31 Å and 2.28 Å, respectively. The ligand also formed a carbon-hydrogen bond with THR-227 at a distance of 2.67 Å. Others include alkyl interaction with VAL-121, π-anion with ASP-124, and π-alkyl with LYS-187 and LEU-257. Distinguishably, the binding interactions of Compound 24 with the receptor were characterized only by electrostatic and hydrophobic interactions, and without hydrogen bond interactions. This may be attributed to the steric hindrances posed by the side chain methyl groups on the phenyl ring and the chloro groups on the pyrrole ring which shields the carbonyl groups from interacting with amino acid residues to form hydrogen bonds. The observed interactions were dominated by the hydrophobic interactions (alkyl type) with the exception of ILE-261 which binds by electrostatic interaction to the π electron system of the benzene ring of compound 24 via its alkyl group. Lastly, the reference drug (pentamidine) was equally docked onto the binding pocket of the pyridoxal kinase receptor in order to provide insight into their binding interaction mode and for validation purpose. Pentamidine interacted with PdxK via two (2) conventional hydrogen bonds with ASN-151 and SER-12 at a distance of 1.95 Å and 2.28 Å, respectively. It equally formed a carbon-hydrogen bond with ASP-231 at a distance of 3.40 Å. Others are electrostatic and hydrophobic interactions including π-alkyl interactions with VAL-19 and LYS-187, π-cation with TYR-85 and π-π stacked with TYR-85.
Structured-based drug design
As reported in Table 11, the predicted pIC50 values of most of the designed compounds (N3, N7-N12) were greater than those of their corresponding template molecules, an evidence of being more biologically active molecules than their templates. Also, all the newly designed compounds showed higher docking scores compared to the template molecules, while only N7 and N12 with binding energy values of − 8.9 and − 8.5 kcal/mol showed better docking scores than that of the reference drug (pentamidine). This is a clear indication of how well these new maleimides would interact with the target enzyme. Because N7 and N12 bind more strongly to PdxK than pentamidine, their interactions with the binding pocket of the protein target shall be discussed further.
N7 was observed from Fig. 9 to bind excellently with the active site of PdxK via six (6) conventional hydrogen bonds, one (1) π-alkyl hydrophobic interaction and one (1) π-anion electrostatic interaction. The carbonyl groups oxygen atoms play a vital role in hydrogen bond formation just as observed with the template molecules. One of the carbonyl groups formed hydrogen bond with ASN-151 and LYS-187 via its oxygen at distance of 2.32 Å and 2.11 Å, respectively. The other carbonyl group formed a hydrogen bond with GLY-228 at a distance of 2.40 Å and 2 hydrogen bonds with THR-229 at a distance of 2.38 Å and 2.40 Å. It also formed an additional hydrogen bond with GLN-258 at a distance of 1.98 Å via a hydroxyl group at the para-position of the outer phenyl ring system. Others include π-alkyl hydrophobic interaction with LEU-198 via the compound’s π electrons of the outer benzene ring, and π-anion electrostatic interaction with ASP-124 via the compound’s π electrons of the pyrrole ring system. The interaction of N12 with the protein target as observed from Fig. 10 was via three (3) conventional hydrogen bonds, three (3) carbon-hydrogen bonds and one (1) π-donor hydrogen bond. The observed hydrophobic interactions include one each of alkyl and π-alkyl interactions, while the electrostatic interactions include one each of π-anion and π-cation interactions. Also, observed was a halogen interaction with the receptor. The observed conventional hydrogen bonds were formed by the interaction of its carbonyl groups oxygen with LYS-187 and THR-229 at distance of 2.06 Å and 2.46 Å, respectively, and between its nitro group oxygen and SER-188 at a distance of 2.64 Å. The carbon-hydrogen bonds were formed by its trifluoromethyl group’s interaction with GLY-230, LYS-187 and THR-229 at interaction distance of 2.71 Å, 3.04 Å and 2.38 Å, respectively. Also observed was a π-donor hydrogen bond between the π-electron of the pyrrole ring system and ASN-151 at a distance of 3.19 Å. The observed hydrophobic interactions include alkyl interaction with LYS-187 and VAL-121, and π-alkyl interaction with LYS-187. Furthermore, the observed electrostatic interactions comprises of π-cation between LYS-187 and its benzene ring π-system, and π-anion between ASP-124 and its pyrrole ring π-system. Therefore, these compounds have demonstrated the potentials to arrest the target receptor (PdxK) an absolutely necessary factor whose inactiveness is dangerous to the viability of the parasite (L. donovani).
Pharmacokinetics properties prediction
According to the Lipinski’s rule for oral bioavailability, a drug molecule is more likely to have poor absorption or permeation when it has hydrogen bond donors (HBD) of greater than 5, hydrogen bond acceptors (HBA) > 10, molecular weight (MW) > 500 and lipophilicity (MLOGP > 4.15 or WLOGP > 5) (Lipinski et al. 2001). Molecules that satisfy at least three out of the four requirements are said to obey the Lipinski’s rule for oral-bioavailability (Lawal et al. 2021).
As seen from Table 13, all the designed molecules perfectly obeyed the Lipinski’s rule by showing no violation. Also, the reported values of topological polar surface area (TPSA) for all molecules were less than 140 Å2, beyond which a molecule may exhibit poor gastrointestinal absorption. Additionally, the synthetic accessibility scores of all tested molecules are in the easy portion (˂ 5.00), indicating their easy laboratory synthesis. The estimated water solubility (Log S) ranges from moderately soluble (N5, N8 and N12) to soluble (the remaining 9 compounds). The Boiled Egg representation in Fig. 11 provides for an intuitive evaluation of passive gastrointestinal absorption and brain barrier permeability as a function of the position of the molecules in the WLOGP against TPSA plot (Daina et al. 2017). As seen from Fig. 11, all the designed molecules were represented in red dots, indicating that they were predicted not to be effluated from the central nervous system by P-glycoprotein. P-glycoprotein is an enzyme which acts as a biological barrier by extruding toxins and xenobiotics, including drugs out of cells. Also, six (6) molecules (N1, N6, N7, N9, N10 and N12) were located in the Boiled Egg’s yolk, meaning that these molecules were predicted to passively permeate through the blood–brain barrier (BBB), while the remaining six (6) molecules were located in the Boiled Egg’s white which is an indication that they were predicted to be passively absorbed by the gastrointestinal tract.
The predicted ADMET properties in Table 14 showed that the human intestinal absorption (HIA) was high (> 70%) for all newly designed compounds. Drug molecules are said to be poorly distributed to the brain through the blood–brain barrier (BBB) and considered as unable to penetrate the central nervous system (CNS), when the values of the logarithmic ratio of brain to plasma drug concentration (logBB) are less than − 1, and the blood–brain permeability-surface area product (logPS) are less than − 3, respectively. Consequently, all the newly designed compounds were predicted to cross the BBB except N7, N10 and N12 with predicted logBB values of less than − 1 (Table 14). Also as predicted in Table 14, all the molecules showed CNS permeability, i.e., logPS > − 3. Furthermore, all the studied molecules except N7 and N11 are substrates of the Cytochrome P450 enzyme (CYP-3A4), an important enzyme for drug metabolism in the body, with none of the compounds inhibiting the enzyme. The degree of drug elimination from the body is measured by the drug’s total clearance, which is within the accepted range for these newly designed compounds. Additionally, the predicted values of maximum recommended tolerated dose (MRTD) for all molecules are included in Table 14. MRTD value of less than or equal to 0.477 log (mg/kg/day) is considered low, and high if greater than 0.477 log (mg/kg/day). The overall predicted drug-likeness and ADMET properties put these molecules on an excellent pharmacokinetics profile, and more so that they are orally bio-available. Therefore, the newly designed molecules are suggested for practical evaluation and/or validation in the laboratory as potential drug candidates for the treatment of leishmaniasis.