Adjustable QSPRs for Prediction of Properties of Long-Chain Substances

The development of quantitative structure property relationships (QSPRs) with good extrapolation capabilities for high carbon number (n(C)) substances in homologous series is considered. Based on the available experimental data, molecular descriptors collinear with a particular property are identified. Among these, the ones whose behavior at the limit n(C) -> infinity is similar to the properties behavior, are eventually used for prediction. A linear QSPR in terms of the selected descriptor with an optional additional correction term which exponentially decays with n(C) can be developed. The confidence level in the property prediction can be adjusted to the quantity, precision, and reliability of the available data. The proposed method has been tested by developing QSPRs for predicting T-C and P-C for several homologous series and T-m for the nalkane series. In all cases, the QSPRs developed represent the available experimental data satisfactorily and converge to theoretically accepted values for n(C) -> infinity. (C) 2010 American Institute of Chemical Engineers AIChE J, 57: 423-433, 2011