SMILES-driven machine learning for high-throughput investigation of anti-corrosion materials

This investigation delves into the viability of the simplified molecular input line entry system (SMILES)-based machine learning (ML) approach as the sole input feature for predicting the corrosion inhibition efficiency (CIE) of pyridine-quinoline compounds to replace various quantum chemical properties (QCP). Employing the molecular access system (MACCS) fingerprint techniques simplifies the processing of molecular structures, enhancing data efficiency. The ML algorithm, notably the gradient boosting (GB) model, showcases superior predictive capabilities, as evidenced by R² and RMSE values of 0.92 and 0.07, respectively. This outcome is akin to predictions employing 20 QCP features, yielding R² and RMSE values of 0.90 and 0.08, respectively. The study substantiates SMILES as a robust single feature for accurate CIE prediction, revealing a moderate correlation between SMILES-represented structures and CIE values. This underscores the effectiveness of SMILES-based ML in assessing corrosion inhibition potential, thereby advancing predictive modeling in corrosion science. Integrating machine learning and SMILES notation presents an efficient approach for evaluating the corrosion inhibition capacity of diverse molecular structures.