QSPR prediction of densities of organic liquids

A general quantitative structure properly relationship (QSPR) treatment of a data set incorporating 303 individual structures (containing C, H, N, O, S, F, Cl, Br and I) from a wide cross section of classes of organic liquids has given an excellent two-parameter correlation for densities (R-2 = 0.9749, s(2) = 0.0021 for rho(20)). The statistical treatment to find the best multi-parameter correlations from subsets of given size within larger sets of molecular descriptors was performed using the scale forward selection method within the CODESSA (comprehensive descriptors for structural and statistical analysis) program. The descriptors involved in the correlation equation that describe the density (rho) are the intrinsic density calculated as the ratio of the molecular mass over the theoretically calculated Van der Waals molecular volume and the total molecular electrostatic interaction per atom in the molecule. It has been demonstrated that with just a two-parameter equation, the densities for compounds that are unknown, unavailable, not easily handled (toxic, odorous, etc.) or not yet synthesized can be predicted with a considerable degree of confidence. (C) 1999 Elsevier Science Ltd. All rights reserved.