Neural Network Based Quantitative Structural Property Relations (QSPRs) for Predicting Boiling Points of Aliphatic Hydrocarbons

A substructural molecular fragment (SMF) method has been developed to model the relationships between the structure of organic molecules and their thermodynamic parameters of complexation or extraction. The method is based on the splitting of a molecule into fragments, and on calculations of their contributions to a given property. It uses two types of fragments: atom/bond sequences and "augmented atoms" (atoms with their nearest neighbors). The SMF approach is tested on physical properties of C2-C9 alkanes (boiling point, molar volume, molar refraction, heat of vaporization, surface tension, melting point, critical temperature, and critical pressures) and on octanol/water partition coefficients. Then, it is applied to the assessment of (i) complexation stability constants of alkali cations with crown ethers and phosphoryl-containing podands, and of beta-cyclodextrins with mono- and 1,4-disubstituted benzenes, and (ii) solvent extraction constants for the complexes of uranyl cation by phosphoryl-containing ligands.