Predicting antibacterial activity of some Curcumin compounds: Fingerprint-based 2D-QSAR models

Abstract

The strong spreads of bacterial infections, caused inter alia by the strong bacterial resistance to drugs, have impelled scientists to look for new antimicrobial agents. A series of Curcumin derivatives with heterocyclic moiety have been evaluated, in terms of MIC (Minimum Inhibitory Concentration), because of their antibacterial activity against Gram-Positive Cocci (Streptococcus aureus) as well as Gram-Negative Bacilli (Escherichia coli) bacterial strains.To investigate the relationship between activities and structures, a 2D-QSAR study is applied to a set of fifty compounds that are split into two sub-groups; the first group contains 29 molecules that exhibited antibacterial activity towards S. aureus bacterium, while the second one includes 27 molecules that displayed the same activity against E. coli.This study was conducted using the kernel-based Partial Least-Square method (KPLS) combined with four non classical descriptors 2D binary fingerprints (linear, dendritic, radial, and MACCS). All forms of the previous fingerprints displayed good statistical indicators: regression correlation coefficient R2> 0.9 and validation coefficient Q2> 0.8.In terms of correlation and validity, the model built from the dendritic fingerprint was chosen as the best KPLS model for both types of bacteria.Relating to the model visualization, Curcumin's ß-diketone moiety and 1,6-heptadiene consistently improve their efficacy against different types of bacteria.The feruloyl moieties substituents in Curcumin Skeleton showed a positive contribution for the hydroxyl radicals, while the methoxy group was unfavorable.Regardless of the type of bacteria, the hydrophobic group produces a favorable interaction for all structures.

Keywords

• Antibacterial activity
• Streptococcus aureus
• Escherichia coli
• QSAR
• Nonlinear regression
• KPLS