Advancing antimicrobial polymer development: a novel database and accelerated design via machine learning

Abstract

The rapid growth of resistant microorganisms has caused serious public health issues and poses great pressure on the current healthcare system. In this environment, the necessity of new antibiotic materials is even more prominent. Antimicrobial polymers are a class of polymers that have the ability to eradicate or impede the proliferation of microbes on their surfaces or within their surrounding environment. The mechanism of action of antibacterial polymers also makes them a perfect fit for medical devices. Despite great growing needs, the design of new antibacterial polymers with desired antimicrobial properties is still challenging. In this work, we present the first open-source database for antimicrobial polymers which consists of 489 entries, with 177 unique polymers exhibiting diverse structures and properties. Multiple predictive models were also designed and trained to classify the antimicrobial properties of these polymers. The best-performing random forest model showed an average accuracy of 86.7% in a 10-fold cross-validation test. We also developed multiple guiding pipelines for the design of novel antimicrobial polymers.