Unlocking potent anti-tuberculosis natural products through structure-activity relationship analysis

Abstract

Tuberculosis (TB) remains a world health problem due to the high number of affected individuals, high mortality rates, prolonged treatment durations, and the increasing prevalence of resistance to commercial TB drugs. The emergence of resistance to anti-TB drugs has necessitated urgent research into drug discovery and development, focusing on novel mechanisms of action against Mycobacterium tuberculosis resistant strains. Natural products, with their remarkable structural diversity and bioactivity, are promising sources for the development of new TB drugs or the identification of potential chemical scaffolds exhibiting potent and novel biological activity with minimal or no cytotoxicity to host cells. This review focuses on potent anti-TB natural products with minimum inhibitory concentration (MIC) values below 5 µg mL^-1 and examines their structure-activity relationship (SAR). Significant characteristics and relevant biological properties of each compound were analysed using a Random Forest, machine learning algorithm, to explore SAR. Using molecular docking, AutoDock Vina was utilised to assess molecular interactions with protein targets, and predictive accuracy was enhanced using the XGBoost machine learning model. These analyses provide insights into the mode of action of these compounds and help identify key structural features contributing to their anti-TB activity. In addition, this review examines the correlation between the potency of selected anti-TB compounds and their cytotoxicity, offering valuable insights for the identification of promising scaffolds in TB drug discovery.

Keywords: Anti-TB scaffold; Natural Products; Structure–activity relationship; Tuberculosis.

Fig. 1

Attractive molecular targets for antitubercular drugs. Reprint with permission from Huszár et al. [1]

Fig. 2

Development of OTB-658 through SAR study

Fig. 3

Inhibitory activities of tryptanrthrin and its derivatives on M. tuberculosis H37Rv and several enzymes [31]

Fig. 4

Overview of the methodology applied in this critical review. Icons sourced from Flaticon under the Basic License (CC3.0, Creative Commons)

Fig. 5

Chemical structures of manzamine alkaloids (1–11)

Fig. 6

Chemical structures of guanidine alkaloids (12–19)

Fig. 7

Chemical structures of pyrrole alkaloids (20–22)

Fig. 8

Chemical structures of pyridoacridine alkaloids (23–31)

Fig. 9

Chemical structures of scalarane sesterterpenes (32–37)

Fig. 10

Chemical structures of benzophenanthridine alkaloids (38–44)

Fig. 11

Chemical structures of bisbenzylisoquinoline alkaloids (45–48)

Fig. 12

Chemical structures of spirobisindole alkaloids (49–52)

Fig. 13

Chemical structures of sesquiterpene lactones (53–59)

Fig. 14

Chemical structures of naphthoquinones (60–67)

Fig. 15

Chemical structures of spirotetronates (68–72)

Fig. 16

Chemical structures of C-glycosylated benz[a]anthraquinones (73–76)

Fig. 17

Chemical structures of diketopiperazines (77–84)

Fig. 18

Chemical structures of preussomerins (85–92)

Fig. 19

Chemical structures of aminolipopeptides (93–95)

Fig. 20

Chemical structure of the newest potent natural anti-TB compounds

Fig. 21

Permutation importance of molecular descriptors, specifically ring system, ring counts and functional groups, to MIC. The impact of each descriptor is represented using mean square error (MSE) as the loss function. A higher feature importance value indicatesa greater impact of the descriptor on the MIC

Fig. 22

The ComplexHeatmap for re-ranking docking score after XGBoost model. PDB ID 1XFC (Alanine racemase alr), 6U7A (Aspartate aminotransferase aspAT), 1KPI (Mycolic acid cyclopropane synthase CmaA2), 6B2Q (Protein kinase A PknA), 5AGS (leucyl-tRNA synthase LeuRS), 3WDB (N-terminal domain of Mycobacterium tuberculosis ClpC1), 1HZP (3-oxoacyl-[acyl-carrier-protein] synthase 3 FabH), 4BFZ (Pantothenate kinase PanK, type 1), 4P8C (Decaprenylphosphoryl-β-d-ribofuranose oxidoreductase DprE1), 4XJO (5′-pyridoxal phosphate (PLP)-dependent aminotransferase (BioA)