Screening of antibacterial compounds with novel structure from the FDA approved drugs using machine learning methods

Area of Science:

• Computational chemistry
• Drug discovery
• Infectious diseases

Background:

• Bacterial infections pose a significant threat, particularly to the elderly, due to compromised immunity.
• The rise of antimicrobial resistance, driven by a lack of new antibiotics, presents a global health crisis.
• Identifying novel antibacterial agents is crucial for effective treatment and public health.

Purpose of the Study:

• To develop and validate predictive models for identifying antibacterial compounds.
• To screen FDA-approved drugs for potential antibacterial activity.
• To discover novel antibacterial small molecules with potential therapeutic applications.

Main Methods:

• Utilized support vector machines and random forest algorithms.
• Trained models on active and inactive antibacterial compound data from the ChEMBL database.
• Screened compounds from the DrugBank database using the developed predictor.

Main Results:

• Both predictive models demonstrated excellent performance (mean accuracy >0.9, mean AUC >0.9).
• Identified 1087 small-molecule drugs with potential antibacterial activity from the DrugBank database.
• Discovered 8 novel antibacterial small-molecule compounds, 5 of which are used in cancer treatment.

Conclusions:

• The developed predictor effectively identifies potential antibacterial compounds from existing drug libraries.
• This strategy offers a novel approach to discovering new antibacterial agents, potentially repurposing existing drugs.
• The identified compounds may offer new therapeutic options for bacterial infections, contributing to lifespan extension.