Abstract Quantitative structure retention relationships (QSRRs) were derived allowing prediction of reversedphase high-performance liquid chromatography relative retention coefficients (RtX/Anq) of 30 anthraquinoids (AQs) on seventeen different octadecyl-bonded silica (C18) stationary phases. The sets of 250 molecular descriptors from DRAGON program and the topological structures of all molecules were calculated. By means of the final variable selection method which is genetic and elimination selection stepwise regression algorithms, two/three optimal descriptors were selected to develop a QSRR models for the prediction of retention coefficients of AQs on each stationary phase with a correlation coefficient between 0.9558 and 0.9811 and a leave-one-out cross-validation correlation coefficient between 0.9415 and 0.9734. The root mean square errors over different seventeen phases are within the range of 0.0511–0.0816. Furthermore, the accuracy of all developed models was confirmed by different internal and external validation. In addition, the non-linear radial basis function-partial least square regression was used to model the QSRRs more accurately. A successful interpretation of the complex relationship between retention coefficients of AQs and the chemical structures is achieved by means of the QSRR methods.
