QSPR MODELING OF MOLAR REFRACTIVITY BASED ON SMILES NOTATION MOLECULAR DESCRIPTORS

This study explores the application of the Monte Carlo optimization method for developing QSPR models that predict the molar refractivity of organic compounds. A dataset of 516 diverse compounds was utilized, randomly split into training and test sets. SMILES notation-based molecular descriptors served as the foundation for model development. A comprehensive assessment of the models’ predictive capabilities and robustness was conducted using various statistical methodologies. The obtained validation metrics demonstrate the strong suitability of the established QSPR models. The best model achieved outstanding statistical quality, with R² values of 0.8985 and 0.8752 for the training and test sets, respectively. Additionally, the Q² values of 0.8971 and 0.8714 further support the model’s generalizability. The Monte Carlo optimization technique effectively identified crucial molecular fragments impacting the molecular descriptors employed in the QSPR modeling process.