Presentation: 2025 ND EPSCoR Annual conference 

October 21, 2025, NDSU Memorial Union, Fargo, North Dakota

Machine Learning Modeling of Fouling-Release Polymeric Coating Materials Applying Combinatorial Mixture-based Descriptors

Machine learning is increasingly used in sustainable materials science to predict properties of mixtures, especially in advanced composite or multi-component materials. Its success depends on high-quality numerical features that reflect structural complexity and synergistic interactions, enabling accurate modeling of biological activity and physicochemical behavior in multi-component systems. To address this, we developed machine learning (ML)-based QSAR models using novel combinatorial mixture-based descriptors (mxb-ML/QSAR). The models were developed on dataset of 40 samples of amphiphilic surface-modifying additives that are used in polymeric coating materials, containing PDMS and poly(SBMA). Target property was the fouling release activity based on Cellulophaga lytica and Navicula incerta microorganisms’ removal. The developed two-variable decision tree (DT) ML model exhibited the best performance in predicting C. lytica bacteria removal at 10 psi, achieving an R2 of 0.98 for the training set and 0.87 for the test set. For C. lytica bacteria removal at 20 psi, the one-variable DT model delivered the best results, with an R2 of 0.89 for the training set and 0.83 for the test set. In predicting N. incerta bacteria removal at 10 psi, the two-variable random forest (RF) model demonstrated the best performance, with an R2 of 0.73 for the training set and 0.72 for the test set. For N. incerta bacteria removal at 20 psi, the two-variable RF model again showed the best performance, achieving an R2 of 0.80 for the training set and 0.82 for the test set. Combinatorial descriptors effectively capture the complex relationships among components, enhancing the predictive performance of machine learning models for fouling release. Additional details of this work will be discussed