Monte Carlo-based QSPR modeling of drop-weight impact sensitivity of energetic materials using hybrid SMILES-graph descriptors.

Abstract

Impact sensitivity under mechanical stimuli is one of the most crucial safety properties of energetic materials. The drop-weight test is a common method for evaluating the impact sensitivity of explosives, with results typically reported as the 50% initiation height (H ₅₀) or drop energy (E ₅₀ (cm)). In this study, three quantitative structure-property relationship (QSPR) models were developed and assessed to predict the logarithm drop-weight impact sensitivity (log E ₅₀) of various types of explosive materials using a Monte Carlo algorithm implemented via CORAL software. The descriptor of correlation weight (DCW) was computed from the SMILES notations of the compounds. The correlation intensity index (CII) and the index of ideality of correlation (IIC) were employed as supplementary parameters for optimizing the correlation weights (CW) utilized in the computation of optimal descriptors. A dataset of 515 explosives or energetic compounds was employed to make QSPR models utilizing hybrid optimal descriptors derived from SMILES and hydrogen filled graph (HFG). The dataset was partitioned into four subsets: active training, passive training, calibration, and validation. The prediction ability of generated models was evaluated by various validation criteria. For the log E ₅₀ endpoint, the models obtained across three independent random data splits achieved an average validation-set R ² of 0.7005 ± 0.0094. An average determination coefficient of 0.7162 ± 0.0129 was obtained for the active training sets. The structural attributes that promote an increase/decrease in log E ₅₀ were extracted and discussed, providing a mechanistic interpretation of the model.