Abstract: Molecular electronegativity distance vector (MEDV), based on the two-dimensional topological structure, consists of 10 parameters which express the interaction between four types of atoms. By introducing the concepts of atomic type and atomic attribute, MEDV is applicable to various molecular structures which contain the multiple bonds and heteroatoms. For six physical properties-boiling point (BP), water solubility ［lg (1/S)］, octanol/water partition coefficient (lgP), molecular total surface area (TSA), molecular volumes (MV) and molecular refractions (MR), the multiple linear regression using the MEDV can provide high-quality QSPR models. The correlation coefficients R are 0.9972,0.9925, 0.9975, 0.9998, 0.9997 and 0.9999,respectively.The standard deviations S are 3.2660, 0.1373, 0.1237, 0.0184, 1.0354 and 0.2084, respectively.To validate the prediction capabilities of these QSPR models for the external prediction set, the leave-one-out (LOO) cross-validation (CV) technique is applied. The corresponding correlation coefficient R²_CVfor the lg(1/S) is 0.9714, and the other R²_CV are all larger than 0.99.The results show that MEDV is an excellent topological index with high property correlation and good structural selectivity.