A newly developed descriptor, three-dimensional holographic vector of atomic interaction field (3D-HoVAIF), was used to describe the chemical structures of purine bases. After variable screening by stepwise multiple regression (SMR) technique, a partial least square (PLS) regression model was built with 3D-HoVAIF. The model was satisfactory comparing to reference since correlation coefficients of molecular modeling (R cum 2 ), cross-validation (Q cum 2 ) and standard deviation of estimation (SD) were 0.966, 0.860 and 0.112, respectively, showing that the model had favorable estimation and prediction capabilities. It was illustrated that information related to retention data of purine bases could preferably be expressed by 3D-HoVAIF with definite physico-chemical meanings and easy structural interpretation for purine bases. It was illustrated that 3D-HoVAIF was to preferably express retention data of purine bases and had definite physicochemical significance. So 3D-HoVAIF was a useful structural expression technique for quantitative structure activity (or property or retention) relationships (QSAR/QSPR/QSRR) study, such as structural characterization and chromatographic retention prediction.