Supercritical Fluid Chromatography (SFC), a high-throughput separation technique, has been widely applied as a promising routine method in pharmaceutical, pesticides, and metabolome analysis in the same way as conventional liquid chromatography and gas chromatography. However, the retention behaviors of many compounds in SFC are not fully investigated. In this study, more than 500 pesticides were analyzed on several polar and nonpolar columns using SFC/MS/MS. Then, partial least squares regression (PLS) was used to explore the retention behaviors of pesticides and construct the quantitative structure-retention relationships under practical gradient elution. The optimized relationships between pesticide structures and pesticide retention were established and validated for predicting power using both internal- and external-validations; hence, several important factors affecting retention of the compounds were identified. In the best case, approximately almost all pesticides in the training set and nearly 80% of pesticides in the external validation set could be predicted with the prediction error of less than 0.5 min. Moreover, the proposed workflow successfully established the local interaction profiles, describing the possible interactions in the 8 studied chromatographic systems, and can be further applied for any groups of compounds under any system conditions.