Page 108 - IMO-2-2

Innovative Medicines & Omics Modeling Aurora-B kinase inhibitors

for high-precision binding mode refinement. Molecules According to Tropsha, a high R value is a necessary but
45
2
were sequentially filtered using HTVS, followed by SP, and not sufficient condition for a reliable QSAR model. This is
ultimately, top-ranked hits were selected for XP docking. further supported by RMSE and Pearson R values. Training
39
Lead molecules were selected based on Glide score set compounds were aligned in the AADRR pharmacophore
rankings and pharmacokinetic properties. model and analyzed using three PLS factors in PHASE. The
QSAR results for AADRR yielded parameters indicating
2.4. MD simulation strong predictive capability (R = 0.971, Q = 0.907, Fit
2
2
Lead molecules with top-ranked Glide scores and value = 403, and SD = 0.154). Therefore, the AADRR
acceptable pharmacokinetic properties were selected model was selected for QSAR analysis. The statistical
for MD simulation studies using the DESMOND parameters of AADRR are summarized in Table 2. The
2
module with OPLS-2005 force field. The protein-ligand R value (0.81) for the test set molecules confirms the
complexes were solvated in an orthorhombic box using a model’s predictive robustness. The spatial arrangement
predefined TIP3P water model. The overall charge was of the five-featured pharmacophore model, along
40
neutralized by adding salt counter-ions. The simulations with inter-feature distance, is shown in Figure 2.
were performed under constant temperature (300 K)
and pressure (1.01325 bar) conditions using the Nose- 3.2. Virtual screening
Hoover thermostat and Martyna-Tobias-Klein barostat In this study, a total of 320,078 hit molecules were retrieved
42
41
methods. The simulations were performed using an from two databases: the NCI database (265,242 compounds)
NPT ensemble by considering the number of atoms, and the Maybridge database (54,836 compounds).
pressure, and timescale. During simulations, the long- The initial screening process utilized “Pharmacophore
range electrostatic interactions were calculated using the Matching” using the AADRR model to select the top 1000
Particle-Mesh-Ewald method. 43,44 compounds. Subsequently, these compounds were further
filtered using Lipinski’s rule of five through the QIKPROP
3. Results and discussion program, resulting in 822 promising compounds with

3.1. QSAR pharmacophore modeling favorable pharmacokinetic (ADME) properties. These 822
compounds were then subjected to a comprehensive docking
A dataset of 40 ligands was randomly selected for the analysis using HTVS, SP, and XP molecular docking.
training set and 18 ligands for the test set. The IC 50
of the compound, defined by the concentration of the To identify key active site residues, known Aurora-B
compound required to inhibit Aurora-B kinase activity by inhibitors – Hesperadin, ZM447439, and VX-680 – were
50%, was computed. The chemical structure, along with initially docked with the human Aurora-B: INCENP
their −logIC values, of the training and test set ligands complex. The results revealed ALA 157 and LYS 106 as
50
are given in Table S1 and S2. The ligands with −logIC important residues for successful binding (Figure S1).
50
higher than 7.7 were considered “active,” those lower The Lipinski’s rule-compliant molecules (N = 822) were
than 6.5 as “inactive,” and those with intermediate values subjected to rapid screening by HTVS, after which the top-
as “moderately active” for the creation of CPHs. After ranked compounds were selected for SP docking. Further,
ligand preparation, scoring hypotheses were evaluated by the first 58 compounds were selected after XP docking. The
keeping the RMSD value below 1.2 Å and a vector score virtual screening workflow is presented in Figure 3. Final
above 0.5. Using the tree-based partition technique, the lead selection from the 58 XP-docked hits was based on Glide
pharmacophore identification resulted in 24 different docking scores, binding interactions, and pharmacokinetic
variant hypotheses. The best hypothesis was selected based properties. Five lead compounds were identified from
on the alignment of site points and vector alignment, the NCI database: NCI ID 695163 (compound 1),
volume overlap, selectivity, number of ligands matched, 327359 (compound 2), 721045 (compound 3), 711797
relative conformational energy, and activity. The hypothesis (compound 4), and 104546 (compound 5). The chemical
with five pharmacophoric – two hydrogen bond acceptors structures of these compounds and the known inhibitors
(A), one hydrogen bond donor (D), and two aromatic rings are presented in Figure 4.
(R), denoted as AADRR – was identified as the best model The primary criterion for the selection of the final five
based on R , SD, F, Q , RMSE, stability, and Pearson R lead compounds was the binding affinity, with higher Glide
2
2
values. The actual and predicted IC values for the training docking scores indicating stronger interactions with the
50
and test set molecules, along with their fitness scores, are Aurora-B active site. The identified interactions between
presented in Table 1. The plot of actual versus predicted the Aurora-B protein and the lead compounds are shown
pIC (−logIC ) for both sets is shown in Figure 1. in Figure 5. The binding interactions of the compounds
50
50
Volume 2 Issue 2 (2025) 102 doi: 10.36922/imo.6547