Identification of chemical features for improved outer membrane permeation in mycobacteria using machine learning

Area of Science:

• Microbiology
• Medicinal Chemistry
• Computational Biology

Background:

• Antibiotic efficacy depends on drug penetration into bacterial cells.
• The unique mycobacterial cell envelope poses a barrier to drug entry, particularly for Mycobacterium tuberculosis.
• Understanding chemical features governing drug permeation is crucial for developing new tuberculosis therapeutics.

Purpose of the Study:

• To profile the mycomembrane permeation of a large library of compounds in Mycobacterium tuberculosis and M. smegmatis.
• To identify chemical features that enhance compound permeation across the mycobacterial outer membrane.
• To establish a framework for designing improved antibiotics targeting Mycobacterium tuberculosis.

Main Methods:

• Utilized the bioorthogonal click chemistry-based PAC-MAN assay to screen 1,572 azide-tagged compounds.
• Employed cheminformatics and machine learning to analyze permeation data and identify key chemical predictors.
• Validated the predictive value of identified features in three distinct molecule series.

Main Results:

• Successfully profiled mycomembrane permeation for 1,572 compounds in both M. tuberculosis and M. smegmatis.
• Identified specific chemical features, including nitrogen-containing aromatic scaffolds like indole, associated with enhanced mycomembrane permeation.
• Demonstrated that these identified chemical features have predictive power for improving drug permeation and anti-mycobacterial activity.

Conclusions:

• Developed a data-driven approach to understand and predict compound permeation through the mycobacterial cell envelope.
• Highlighted nitrogen-containing aromatic scaffolds as promising features for enhancing antibiotic efficacy against Mycobacterium tuberculosis.
• Provided a rational framework for the design of novel therapeutics with improved cell penetration and whole-cell activity against tuberculosis.