Capturing antibacterial natural products with in silico techniques

Area of Science:

• Medicinal Chemistry
• Pharmacology
• Computational Drug Discovery

Background:

• The urgent need for novel, cost-effective antibacterial therapeutics drives research into alternative drug sources.
• Natural products represent a vast, underexplored reservoir of potential antibacterial agents.

Purpose of the Study:

• To develop and validate a computational model for indexing natural products based on their predicted antibacterial bioactivity.
• To facilitate the identification of novel antibacterial drug candidates from natural sources.

Main Methods:

• Utilized a dataset of 628 antibacterial drugs (active domain) and 2,892 natural products (inactive domain).
• Employed an iterative stochastic elimination algorithm to identify 36 unique filters for model construction.
• Developed a robust prediction model with an area under the curve (AUC) of 0.957.

Main Results:

• The model achieved high discriminative power, capturing 72% of known antibacterial drugs in the top 1% of screened substances (enrichment factor of 72).
• Identified 10 natural products as high-scoring antibacterial drug candidates.
• Two of these compounds, caffeine and ricinine, were confirmed to possess antibacterial activity, while 8 await experimental validation.

Conclusions:

• The developed indexing model is efficient and rapid for virtual screening of large chemical databases.
• This approach can significantly accelerate the drug discovery and development process for new antibacterial agents.
• The model holds promise for identifying novel, cost-effective antibacterial drug candidates from natural product libraries.