Abstract
The CORAL software was used to build up quantitative structure-property relationships (QSPRs) for the retention characteristics of 93 derivatives of three groups of heterocyclic compounds: 2-phenyl-1,3-benzoxazoles, 4-benzylsulfanylpyridines, and benzoxazines. The QSPRs are one-variable models based on the optimal descriptors calculated from the molecular structure represented by simplified molecular input-line entry systems (SMILES).
Each symbol (or two undivided symbols) of SMILES is characterized by correlation weight. The optimal descriptor is the sum of the correlation weights.
The numerical data on the correlation weights were calculated with the Monte Carlo method by the manner which provides best correlation between endpoint and optimal descriptor for the calibration set. The predictive ability of the model is checked with the validation set (compounds invisible during building up of the model).
The approach has been checked with three random splits into the training, calibration, and validation sets: all models have apparent predictive potential. The mechanistic interpretation of the molecular features extracted from SMILES as the promoters of increase or decrease of examined endpoints is suggested.