A Combined Bayesian and Similarity-Based Approach for Predicting E. coli Biofilm Inhibition by Phenolic Natural...

Area of Science:

• Natural Product Chemistry
• Computational Chemistry
• Microbiology

Background:

• Screening natural compounds for antibiofilm activity is challenging due to chemical diversity, limited availability, and costly lab processes.
• In silico prediction methods are valuable but rely heavily on the quality and quantity of experimental data.

Purpose of the Study:

• To develop and validate a predictive model for antibiofilm activity of phenolic natural compounds.
• To improve the efficiency of identifying potential antibiofilm agents.

Main Methods:

• Experimental assessment of antibiofilm activity for 320 phenolic compounds against Escherichia coli.
• Training a Bayesian logistic regression model combined with a similarity-based method.
• Validation of the predictive model using independent experiments with resistant E. coli strains.

Main Results:

• The predictive model successfully identified antibiofilm activity compared to baseline accuracy.
• The prediction-based substance selection increased the percentage of active phenolic compounds from 6.0% to 34.8%.

Conclusions:

• The developed in silico model effectively predicts antibiofilm effects of phenolic compounds.
• This approach significantly enhances the discovery rate of natural antibiofilm agents, overcoming limitations of traditional screening.