A Simple and Reliable Model for Predicting Impact Sensitivity of Energetic Cocrystals Based on their Structural Properties

In this study, multiple linear regression (MLR) was used as a safe, fast, inexpensive and common method to predict impact sensitivity of energetic cocrystals. For this purpose, the data related to 28 of energetic cocrystal compounds were collected from different reliable references. Then various descriptors such as topological, structural, sub structural and … descriptors were calculated using DRAGON software. The results showed that optimum elemental composition and number of nine-member rings in the cocrystal structure are the most effective descriptors in this model. Determination coefficient of the new model was 0.954 for 22 members training set and root mean square deviation (RMSD) and average absolute deviation (AAD) were calculated as 4.55 and 3.53 J, respectively. Also, RMSD and AAD for six members test set as external validation were derived 4.48 and 3.34 respectively. Also, RMSD and AAD were 4.36 and 3.95 respectively for cross validation of derived model. For investigating of predictability of the new model were used two cross validation method as internal validation. The results showed that the values (Q2LOO=0.927 and Q2LMO=0.958). The Roy et. al computations were also used for confirmation of internal validation.