How do cyclic antibiotics with activity against Gram-negative bacteria permeate membranes? A machine learning...

Area of Science:

• Microbiology
• Biochemistry
• Computational Biology

Background:

• Antibiotics must overcome bacterial membrane barriers to be effective.
• Gram-negative bacteria pose a significant public health threat due to their resistance.
• Cyclic antibiotics, often of bacterial origin, show promise against these pathogens.

Purpose of the Study:

• To identify common mechanisms of interaction between cyclic antibiotics and bacterial lipid membranes.
• To investigate the role of negative Gaussian curvature (NGC) in cyclic antibiotic activity.
• To evaluate a machine-learning classifier's ability to predict NGC induction by cyclic antibiotics.

Main Methods:

• Analysis of diverse cyclic antibiotics (bactenecin, polymyxin B, octapeptin, capreomycin, Kirshenbaum peptoids).
• Investigation of their ability to induce negative Gaussian curvature (NGC) in bacterial membranes.
• Application of a modified machine-learning classifier trained on antimicrobial peptides (AMPs) to predict NGC induction.

Main Results:

• Most cyclic antibiotics examined induce negative Gaussian curvature (NGC) in bacterial membranes.
• The machine-learning classifier accurately predicted NGC induction for bactenecin and polymyxin B, but not capreomycin.
• The classifier successfully replicated structure-activity relationships observed in polymyxin B.

Conclusions:

• Negative Gaussian curvature (NGC) is a common feature in the membrane interaction of active cyclic antibiotics.
• Machine-learning models can predict the membrane-disrupting potential of cyclic antibiotics.
• Shared sequence design principles may exist between cyclic antibiotics and linear antimicrobial peptides (AMPs).