Prediction of density of energetic cocrystals based on QSPR modeling using artificial neural network

Among the most important factors affecting the destructive power of explosive materials is density. In order to correlate the molecular structure of energetic cocrystals (ECs) with their density (ρ), a quantitative structure-property relationship (QSPR) study was undertaken. An artificial neural network (ANN) model was developed to predict the density of cocrystals by using three out of more than 1600 molecular descriptors, computed by Dragon software, as input variables. The complete set of 26 ECs was randomly divided into a training set of 16, a test set of 5, and a validation set of 5 compounds. Also, multiple linear regression (MLR) analysis was utilized to build a linear model by using the same descriptors. Correlation coefficient (R2) of the ANN and MLR models (for the whole dataset) was 0.9716 and 0.9309, respectively. The ANN model was further investigated, and average absolute relative deviation for the complete dataset was 2.48%, indicating good accuracy and reliability of the model.