Abstract
Here we present a novel and simple approach for predicting retention indices of polycyclic aromatic sulfur heterocycles on (5%-Phenyl)-methylpolysiloxane stationary phase using a new set of structural descriptors based on substituent positions. The prediction method modeled correlations between retention indices and structural parameters for alkyl substituted benzothiophenes and dibenzothiophenes. Five models were created: two for uni- and di- substituted benzothiophenes homologous series, and three for the following non-homologous series: (i) uni- to tri- substituted benzothiophenes; (ii) uni- to tetra-substituted dibenzothiophenes and (iii) isomers of benzo[b]naphtha[2,1-d]thiophene and benzo[b]naphtha[1,2-d]thiophene. Retention was highly correlated with structural parameters, and all models had high correlation coefficients, low root mean-squared errors, and low mean absolute percentage errors for the homologous series. For non-homologous models, R-values were 0.998 for dibenzothiophenes, 0.995 for benzothiophenes, and 0.991 for benzo[b]naphtha[2,1-d]thiophene and benzo [b] naphtha[1,2-d]thiophene, lower than for homologous series. Our approach represents a simple strategy for the prediction of experimental retention indices values without the need for state-of-the-art technology.