Linear QSPRs for predicting pure compound properties in homologous series

Abstract

Abstract Linear QSPRs, containing 1 through 4 descriptors, are developed for predicting the normal boiling temperature, melting point temperature, and critical properties for the n‐alkane, 1‐alkene, n‐alkylbenzene, 1‐alcohol, and alkanoic monocarboxylic acid homologous series. It has been shown that property values for which experimental data are available can be predicted within experimental error level (with very few and very small exceptions), irrespective of whether interpolation or extrapolation is involved. Property values for which predicted literature data are available can be matched within the reported “reliability” level, even when extrapolation is carried out from very small training sets containing experimental data. Thus, the linear QSPRs developed represent well the nonlinear variation of the particular property with the carbon number, and increase the confidence in the values predicted when extrapolation is involved. © 2008 American Institute of Chemical Engineers AIChE J, 54: 978–990, 2008