Abstract: Quantitative structure-property relationships (QSPR) have been demonstrated to be a powerful tool in chromatography. QSPR have been used to obtain simple models to explain and predict the chromatographic behavior of various classes of compounds. The study of quantitative structure and retention index relationship (QSRR) is an important subject in chromatographic field. One hundred twenty-seven topological descriptors of 207 methylalkane structures are calculated. GAPLS method, which is a variable selection method combining with genetic algorithms (GA), back stepwise and partial least squares (PLS), is introduced in the variable selection of quantitative structure gas chromatographic (GC) retention index relationship. Seven topological descriptors are selected from 127 topological descriptors by GAPLS method to build QSRR model with high regression quality: squared correlation coefficient (R2) of 0.99998, standard deviation (S) of 2.88. The error of the model is similar to the experimental error. The validation of the model is checked by leave-one-out cross-validation technique. The result of leave-one-out cross-validation indicates that the built model is reliable and stable with high prediction quality, such as squared correlation coefficient of leave-one-out (R2cv) of 0.99997 and standard deviation of leave-one-out predictions (Scv) of 2.95. A successful interpretation of the complex relationship between GC retention indexes of methylalkanes and the chemical structure is achieved using QSPR method. The seven variables in the model are also rationally interpreted, which indicates methylalkane retention index are precisely represented by topological descriptors.

Key words: chromatographic retention indexes, methylalkanes , quantitative structure and retention index relationships, GAPLS method