Machine learning-enabled virtual screening indicates the anti-tuberculosis activity of aldoxorubicin and quarfloxin with verification by molecular docking, molecular dynamics simulations, and biological evaluations

Abstract

Drug resistance in Mycobacterium tuberculosis (Mtb) is a significant challenge in the control and treatment of tuberculosis, making efforts to combat the spread of this global health burden more difficult. To accelerate anti-tuberculosis drug discovery, repurposing clinically approved or investigational drugs for the treatment of tuberculosis by computational methods has become an attractive strategy. In this study, we developed a virtual screening workflow that combines multiple machine learning and deep learning models, and 11 576 compounds extracted from the DrugBank database were screened against Mtb. Our screening method produced satisfactory predictions on three data-splitting settings, with the top predicted bioactive compounds all known antibacterial or anti-TB drugs. To further identify and evaluate drugs with repurposing potential in TB therapy, 15 screened potential compounds were selected for subsequent computational and experimental evaluations, out of which aldoxorubicin and quarfloxin showed potent inhibition of Mtb strain H37Rv, with minimal inhibitory concentrations of 4.16 and 20.67 μM/mL, respectively. More inspiringly, these two compounds also showed antibacterial activity against multidrug-resistant TB isolates and exhibited strong antimicrobial activity against Mtb. Furthermore, molecular docking, molecular dynamics simulation, and the surface plasmon resonance experiments validated the direct binding of the two compounds to Mtb DNA gyrase. In summary, our effective comprehensive virtual screening workflow successfully repurposed two novel drugs (aldoxorubicin and quarfloxin) as promising anti-Mtb candidates. The verification results provide useful information for the further development and clinical verification of anti-TB drugs.

Keywords: Mycobacterium tuberculosis; antitubercular; drug repurposing; ligand-based virtual screening; machine learning.

Graphical Abstract

Figure 1

The overall workflow of this study.

Figure 2

QED property distributions of the compounds from the ChEMBL dataset and their correlations with MIC. (A) Molecular weight (MW), (B) AlogP, (C) number of hydrogen bond acceptors (HBA), (D) number of hydrogen bond donors (HBD), (E) polar surface area (PSA), (F) number of rotatable bonds (ROTB), (G) number of aromatic rings (AROM), and (H) number of alert structures (ALERTS).

Figure 3

Chemical structure space analysis. (A) Representative molecular scaffolds in the top 5 bioactive molecular clusters. (B) T-SNE plot data and common chemical substructures in three representative bioactive molecular clusters in the ChEMBL dataset (red dots: bioactive molecules; blue dots: bioinactive molecules). (C) T-SNE plot data from the ChEMBL and DrugBank datasets.

Figure 4

(A) Scatter plot showing the prediction results of the ENSEM on the ChEMBL internal validation dataset with random splitting (x-axis: actual MIC values; y-axis: EnSEM predicted MIC values). (B) Scatter plot showing the prediction results of the ensemble model on the Lane et al. external validation dataset (x-axis: actual MIC values; y-axis: EnSEM predicted MIC values). (C) Rank-ordered predicted bioactivities for DrugBank molecules. (D) Average bioactivities of the overall, top 300 ranked, and bottom 300 ranked molecules in the ChEMBL internal validation dataset. (E) Average bioactivities of the overall, top 30 ranked, and bottom 30 ranked molecules in the Lane et al. external validation dataset. (F) Average predicted bioactivities of the overall, top 300 ranked, and bottom 300 ranked molecules in the DrugBank repurposing dataset.

Figure 5

Time–kill curves of aldoxorubicin (A), quarfloxin (B), and INH (C) against Mtb H37Rv. Antibiotic concentrations are presented as different symbols. The INH was used as a positive control.

Figure 6

The intracellular survival rate of Mtb H37Rv after aldoxorubicin and quarfloxin exposure. (A) Aldoxorubicin group: infected macrophages treated with aldoxorubicin at 1× MIC, 2× MIC, and 4× MIC. (B) Quarfloxin group: infected macrophages treated with quarfloxin at 1× MIC, 2× MIC, and 4× MIC. The INH group served as the positive control. ^***P < .001.

Figure 7

The protein–ligand binding modes of two identified ligands (aldoxorubicin and quarfloxin) and reference ligand (evybactin). (A) The global protein–ligand complexes of Mtb DNA gyrase (PDB ID: 7UGW) and ligands. (B) The local protein–ligand complexes and binding modes of Mtb DNA gyrase and three ligands, with protein hydrophobic surface and binding residues shown. (C) The ligand-centered binding mode profiles for three ligands.

Figure 8

The MD simulation analysis results for Mtb DNA gyrase monomer and three protein–ligand complexes (Mtb DNA gyrase–aldoxorubicin, –quarfloxin, and –evybactin). (A) Time evolution of RMSD during MD production; (B) RMSF for each residues in Mtb DNA gyrase during MD production. The 1–46 numbered residues representing DNA molecule were removed. (C) Time evolution of hydrogen bonds formed between aldoxorubicin and protein. (D) Time evolution of hydrogen bonds formed between quarfloxin and protein. (E) Time evolution of hydrogen bonds formed between evybactin and protein.

Figure 9

SPR analysis of aldoxorubicin and quarfloxin with DNA gyrase. (A) Evaluation of interaction between aldoxorubicin and GyrA. (B) Evaluation of interaction between aldoxorubicin and GyrB. (C) Evaluation of interaction between quarfloxin and GyrA. (D) Evaluation of interaction between quarfloxin and GyrB.