Background: It has been demonstrated a variety of condenced phase macroscopic properties that reflect non-covalent interactions can be expressed analytically in terms of statistically defined quantities that characterize molecular surface electrostatic potentials [1]. Methods: Quantitative structure-(property/retention) relationship (QS(P/R)R) models have been developed to predict properties (solubility, boiling point, enthalpies of formation, heats of atomization, n-octanol/water partition coefficient, vapor pressure) and retention (on four stationary phase) of alkyl benzene derivatives based on descriptors resulted from computed electrostatic potential and average local ionization energy on molecular surface [2,3]. First all molecules structures optimizaed with HF/STO-3G in Gaussian98 software then these optimized moleules were used to compute electrostatic potential on molecular surfae at B3LYP/6-31G* level in Gaussian98 software. Computed electrostatice potential and average local ionization energy on molecular surface in 106 points were used to calculate 21 statistical parameters (GIPF descriptors) for construction QSPR and QSRR models [2,3]. Elimination stepwise selection procedure was used to select descriptors to create multiple linear regression (MLR) models [4]. Results: Square correlation coefficient (R2) and root mean sqaure errors (RMSEs) of models are between 0.9261-0.9999 and 0.08-69-36, respectively. The accuracy of all the developed models was confirmed using leave-one-out (LOO) cross validation and Y-randomization internal validation metgods [5]. Conclusion: Created models predict properties and retention alkyl benzene derivatives and demonstrated these properties depend on interactions that electrostatic in nature.
