Prediction of Water Contact Angle with UMAP Visualization for Methacrylate‐Based Copolymer Surfaces and Model Applicability

ABSTRACT

Predicting the surface wettability of copolymers remains challenging because interfacial properties are governed not only by bulk composition but also by physicochemical phenomena such as surface segregation and molecular reorientation. In this study, machine learning models were developed to predict the water contact angle (WCA) of dip‐coated methyl methacrylate (MMA)‐based copolymer surfaces using composition‐weighted molecular descriptors. A dataset of 135 copolymers prepared from 82 monomers under unified experimental conditions was constructed. Linear regression provided chemically interpretable relationships between molecular descriptors and WCA, whereas XGBoost achieved higher predictive accuracy. Chemical‐space visualization using uniform manifold approximation and projection (UMAP) enabled assessment of model applicability by distinguishing interpolation and extrapolation regions. Validation using newly synthesized copolymers showed that predictions were reliable within well‐represented chemical space, whereas systematic deviations were consistent with physicochemical phenomena such as surface segregation and molecular reorganization. These findings demonstrate that combining machine learning with chemical‐space analysis not only enables assessment of model applicability but also provides insight into the physicochemical limitations of descriptor based models for polymer surface design.