Linear Binary Classifier to Predict Bacterial Biofilm Formation on Polyacrylates.

Area of Science:

• Biomaterials Science
• Computational Chemistry
• Infectious Disease Research

Background:

• Antimicrobial resistance necessitates novel strategies to prevent medical device-associated infections.
• Biofilm formation on materials is a primary cause of persistent infections.
• Machine learning (ML) shows promise in predicting bacterial adhesion to materials.

Purpose of the Study:

• To develop interpretable ML models for predicting bacterial adhesion to polyacrylates.
• To identify key material properties influencing bacterial attachment.
• To guide the rational design of pathogen-resistant coatings.

Main Methods:

• Utilized interpretable mass spectral molecular ions and chemoinformatic descriptors.
• Employed a linear binary classification model for bacterial attachment prediction.
• Analyzed feature importance and correlated it with chemical descriptors.

Main Results:

• Developed predictive models for the attachment of *Pseudomonas aeruginosa* and *Staphylococcus aureus*.
• Identified interpretable chemoinformatic descriptors that robustly predict bacterial attachment.
• Derived rules elucidating structure-function relationships in material-pathogen interactions.

Conclusions:

• Interpretable ML models offer improved guidance for designing anti-attachment materials.
• Chemoinformatic descriptors can effectively predict bacterial adhesion to polyacrylates.
• This approach facilitates the identification of novel materials to combat nosocomial infections.