A Computer-Aided Drug Design Approach to Predict Marine Drug-Like Leads for SARS-CoV-2 Main Protease Inhibition

Abstract

The investigation of marine natural products (MNPs) as key resources for the discovery of drugs to mitigate the COVID-19 pandemic is a developing field. In this work, computer-aided drug design (CADD) approaches comprising ligand- and structure-based methods were explored for predicting SARS-CoV-2 main protease (M^pro) inhibitors. The CADD ligand-based method used a quantitative structure-activity relationship (QSAR) classification model that was built using 5276 organic molecules extracted from the ChEMBL database with SARS-CoV-2 screening data. The best model achieved an overall predictive accuracy of up to 67% for an external and internal validation using test and training sets. Moreover, based on the best QSAR model, a virtual screening campaign was carried out using 11,162 MNPs retrieved from the Reaxys^® database, 7 in-house MNPs obtained from marine-derived actinomycetes by the team, and 14 MNPs that are currently in the clinical pipeline. All the MNPs from the virtual screening libraries that were predicted as belonging to class A were selected for the CADD structure-based method. In the CADD structure-based approach, the 494 MNPs selected by the QSAR approach were screened by molecular docking against M^pro enzyme. A list of virtual screening hits comprising fifteen MNPs was assented by establishing several limits in this CADD approach, and five MNPs were proposed as the most promising marine drug-like leads as SARS-CoV-2 M^pro inhibitors, a benzo[f]pyrano[4,3-b]chromene, notoamide I, emindole SB beta-mannoside, and two bromoindole derivatives.

Keywords: actinomycetes; drug discovery; machine learning (ML) techniques; main protease enzyme (Mpro); marine natural products (MNPs); molecular docking; quantitative structure–activity relationship (QSAR); severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2); virtual screening.

Figure 1

Chemical structures of the five molecules in test set 2 that were predicted as class A with the confidence value (3) and Prob_A ≥ 0.5.

Figure 2

Chemical structure of quizartinib from the test set 2 that was predicted as class A with the confidence value (3) and Prob_A of 0.46, which is active in the SARS-CoV-2 CPE assay.

Figure 3

Chemical structure of vorapaxar from the test set 2 that was predicted as class A with the confidence value (1) and Prob_A of 0.17, which is active in the SARS-CoV-2 CPE assay.

Figure 4

The twenty most important MACCS FPs and 1D&2D descriptors selected in RF classification models. Where: ¹ FPs with a nitrogen atom; ² FPs that can have at least one nitrogen atom as A (any atom) or as Q (any non-C or non-H atom); ³ Autocorrelation descriptors; ⁴ Topological descriptors; ⁵ Atom-type descriptors.

Figure 5

Interaction profiles of the best-docked poses for the five hits using the set search space coordinates, in which a higher affinity was accomplished (a) X: −36.149 Y: −3.796 Z: 45.045; or (b) X: −12.806 Y: 18.646 Z: 65.607.