First QSPR models to predict the thermal stability of potential self-reactive substances

Self-reactive substances are unstable chemical which can easily decompose and may lead to explosion in transport, storage, or process situations. For this reason, their thermal stability properties required assess possible safety issues for classification purpose. In study, the first quantitative structure–property relationships (QSPR) dedicated class of compounds were developed predict heat decomposition self-reactive from molecular structures. The database used develop validate models was issued a experimental campaign on 50 samples using differential scanning calorimetry homogeneous conditions. QSPR derived GA-MLR methods (using genetic algorithm multi-linear regressions) descriptors calculated by Dragon software based two types inputs: 3D structures determined density functional theory (DFT), allowing access three-dimensional descriptors, SMILES codes, favoring simpler models, requiring no preliminary quantum calculations. All respected OECD validation guidelines regulatory acceptability models. They tested internal external tests applicability domains defined analyzed.