Potentials of Machine Learning in Predicting Key Features of Synthetic Antimicrobial Polymers.

Area of Science:

• Polymer Chemistry
• Computational Chemistry
• Antimicrobial Research

Background:

• Rising antimicrobial resistance necessitates novel therapeutic strategies.
• Synthetic antimicrobial polymers (SAMPs) offer tunable alternatives to host-defense peptides.
• Understanding structure-activity relationships is crucial for designing effective SAMPs.

Purpose of the Study:

• To use machine learning (ML) to explore structure-activity relationships in a polyacrylamide library.
• To identify key polymer features influencing antimicrobial activity and hemagglutination.
• To evaluate ML algorithms for predicting SAMP performance and toxicity.

Main Methods:

• Systematic synthesis of 23 polyacrylamide designs varying in side-chain chemistry, chain length, cationic amine ratio, and architecture.
• Evaluation of minimum inhibitory concentrations (MIC) against *Pseudomonas aeruginosa* and *Staphylococcus aureus* strains.
• Assessment of red blood cell agglutination for hemagglutination.
• Application and comparison of regression random forest and gradient boosting ML algorithms.
• Utilized Shapley additive explanations (SHAP) for feature importance analysis.

Main Results:

• Gradient boosting regression demonstrated superior predictive power with low Root Mean Square Error (RMSE) for MIC and hemagglutination.
• Feature importance analysis revealed that the type and percentage of cationic monomer are critical for SAMP efficacy.
• Three polymer designs showed low MIC values, and five exhibited low hemagglutination.
• Gradient boosting models achieved RMSE values of 20, 6, 13, and 12 μg/ml for four bacterial strains and 1 μg/ml for hemagglutination.

Conclusions:

• Boosting-ensemble ML methods, particularly gradient boosting, provide robust predictive capabilities for SAMP design.
• ML can effectively forecast the potency and toxicity of novel synthetic antimicrobial polymers.
• These findings support the use of ML in accelerating the discovery of new antimicrobial agents.