Ninety sulfur compounds consisting of mercaptans, sulfides and thiophenes, were identified in a fluid-catalytic-cracking light cycle oil using gas chromatography with atomic emission detection. Their retention times and indices were correlated with molecular descriptors generated from their molecular structures. The best seven- and eight-parameter multi-linear regression models showed good predictive ability. The descriptors involved in the models reflect the geometrical, topological, and electronic properties of the molecules, related to the interactions between the solute and the stationary phase. For the 34 thiophenic sulfur compounds (benzothiophenes and dibenzothiophenes) of most interest in petroleum processing, another two five-parameter multi-linear models were developed for retention times and indices with standard errors s = 0.61 and 1.63, respectively. Such models for retention times and/or indices can be used for identification of unknown chromatographic peaks by matching their retention times or indices with those of sulfur compounds of known molecular structure when the corresponding chemical standards are unavailable. (C) 2004 Elsevier B.V. All rights reserved.