Leveraging Molecular Descriptors and Explainable Machine Learning for Monomer Conversion Prediction in Photoinduced Electron Transfer-Reversible Addition- Fragmentation Chain Transfer Polymerization

Abstract This study introduces a molecular descriptor-based machine learning (ML) architecture for predicting monomer conversion in photoinduced electron transfer-reversible addition-fragmentation chain transfer (PET-RAFT) polymerization systems. Unlike conventional polymer informatics approaches that treat polymers as single entities or rely on one-hot encoding of reaction components, we decompose the PET-RAFT system into its molecular building blocks: monomer, RAFT agent, and photocatalyst, encoding each component separately through SMILES-derived descriptors supplemented with thermodynamic parameters. Using a 152 PET-RAFT reactions dataset, we systematically trained (with five-fold cross-validation) and tested 10 ML algorithms. CatBoost demonstrated superior stability across CV folds and was identified as the top-performer for monomer conversion prediction (R² = 0.84; RMSE = 10.04 pps; MAE = 8.16 pps). Through SHapley Additive exPlanations (SHAP) analysis, mechanistically interpretable structure-property relationships were revealed that monomer topological complexity, electronic polarization, and molecular weight collectively account for over 60% of the model’s predictive power. External validation demonstrated CatBoost’s generalization ability to unseen (meth)acrylates/ (meth)acrylamides, achieving a mean absolute error (MAE) of 8.03, suggesting improved performance compared to training (9.62 ± 1.71 pps). Furthermore, this explainable, descriptor-based ML approach bridges mechanistic understanding with predictive modelling, providing experimentally actionable hypotheses for rational polymer design while maintaining the interpretability for scientific insights.