Computational modelling studies of some 1,3-thiazine derivatives as anti-influenza inhibitors targeting H1N1 neuraminidase via 2D-QSAR, 3D-QSAR, molecular docking, and ADMET predictions

Abstract

Background: Influenza virus disease remains one of the most contagious diseases that aided the deaths of many patients, especially in this COVID-19 pandemic era. Recent discoveries have shown that the high prevalence of influenza and SARS-CoV-2 coinfection can rapidly increase the death rate of patients. Hence, it became necessary to search for more potent inhibitors for influenza disease therapy. The present study utilized some computational modeling concepts such as 2D-QSAR, 3D-QSAR, molecular docking simulation, and ADMET predictions of some 1,3-thiazine derivatives as inhibitors of influenza neuraminidase (NA).

Results: The 2D-QSAR modeling results showed GFA-MLR ( $R_{\text{train}}^2$ = 0.9192, Q ² = 0.8767, R ² _adj = 0.8991, RMSE = 0.0959, $R_{\text{test}}^2$ = 0.8943, $R_{\text{pred}}^2$ = 0.7745) and GFA-ANN ( $R_{\text{train}}^2$ = 0.9227, Q ² = 0.9212, RMSE = 0.0940, $R_{\text{test}}^2$ = 0.8831, $R_{\text{pred}}^2$ = 0.7763) models with the computed descriptors as ATS7s, SpMax5_Bhv, nHBint6, and TDB9m for predicting the NA inhibitory activities of compounds which have passed the global criteria of accepting QSAR model. The 3D-QSAR modeling was carried out based on the comparative molecular field analysis (CoMFA) and comparative similarity indices analysis (CoMSIA). The CoMFA_ES ( $R_{\text{train}}^2$ = 0.9620, Q ² = 0.643) and CoMSIA_SED ( $R_{\text{train}}^2$ = 0.8770, Q ² = 0.702) models were found to also have good and reliable predicting ability. The compounds were also virtually screened based on their binding scores via molecular docking simulations with the active site of the NA (H1N1) target receptor which also confirms their resilient potency. Four potential lead compounds (4, 7, 14, and 15) with the relatively high inhibitory rate (> 50%) and docking (> - 6.3 kcal/mol) scores were identified as the possible lead candidates for in silico exploration of improved anti-influenza agents.

Conclusion: The drug-likeness and ADMET predictions of the lead compounds revealed non-violation of Lipinski's rule and good pharmacokinetic profiles as important guidelines for rational drug design. Hence, the outcome of this research set a course for the in silico design and exploration of novel NA inhibitors with improved potency.

Keywords: Binding score; Modeling; Neuraminidase; Receptor; Residual interaction.

Fig. 1

Optimized structures (A) structure of compound 4, (B) alignment and superposition of the dataset compounds (capped sticks model)

Fig. 2

Schematic representation of the GFA-ANN (4-5-1) architecture

Fig. 3

Mean effect plot of the model descriptors

Fig. 4

Plot of standardized residuals versus experimental NA activity, A GFA-MLR model, B GFA-ANN model

Fig. 5

Scatter plot of the standardized residuals of the GFA-ANN model against leverage scores (Williams plot)

Fig. 6

Scatter plot of predicted against experimental NA inhibitory activity: A CoMFA_SE model, B CoMSIA_SED model

Fig. 7

3D fields of the CoMFA model for the most active compound 4. A Green areas depict desirable steric bulk, while yellow areas disfavor steric bulk, B electrostatic contour map where blue regions favor positive charge and red regions favor negative charge

Fig. 8

3D fields contribution of the CoMSIA_EAD model for the most active compound 4. A Magenta contours represent regions for desirable hydrogen bond acceptors, while red areas represent undesirable acceptors, B electrostatic contour map where blue regions favor positive charge and red regions favors negative charge, C cyan contours represent areas for desirable hydrogen bond donors, while purple areas represent undesirable donors

Fig. 9

General description of the 3D QSAR analysis

Fig. 10

3D docking view of compound 4 with the H1N1 neuraminidase receptor (PDB: 3TI6). A The best pose of compound 4, B residual interaction of compound 4-complex, C 3D hydrogen bond surfaces around the ligand, D 2D residual interaction of 4-complex

Fig. 11

3D docking view of compound 15 with the H1N1 neuraminidase receptor (PDB: 3TI6). A The best pose of compound 15, B residual interaction of compound 15-complex, C 3D hydrogen bond surfaces around the ligand, D 2D residual interaction of compound 15-complex

Fig. 12

Physicochemical radar chart of the lead compounds in the dataset