Development of quantitative structure property relationships for predicting the melting point of energetic materials

The accurate prediction of the melting temperature of organic compounds is a significant problem that has eluded researchers for many years. The most common approach used to develop predictive models entails the derivation of quantitative structure-property relationships (QSPRs), which are multivariate linear relationships between calculated quantities that are descriptors of molecular or electronic features and a property of interest. In this report the derivation of QSPRs to predict melting temperatures of energetic materials based on descriptors calculated using the AM1 semiempirical quantum mechanical method are described. In total, the melting points and experimental crystal structures of 148 energetic materials were analyzed. Principal components analysis was performed in order to assess the relative importance and roles of the descriptors in our QSPR models. Also described are the results of k means cluster analysis, performed in order to identify natural groupings within our study set of structures. The QSPR models resulting from these analyses gave training set R(2) values of 0.6085 (RMSE = ± 15.7 °C) and 0.7468 (RMSE = ± 13.2 °C). The test sets for these clusters had R(2) values of 0.9428 (RMSE = ± 7.0 °C) and 0.8974 (RMSE = ± 8.8 °C), respectively. These models are among the best melting point QSPRs yet published for energetic materials.