Prediction of the upper flammability limits of organic compounds from molecular structures

A quantitative structure-property relationship (QSPR) study is performed to develop mathematical models for prediction of the upper flammability limits (UFL) of organic compounds from their molecular structures. The structural features of the compounds are numerically represented by various kinds of calculated molecular descriptors such as topological, charge, and geometric descriptors. The genetic algorithm combined with multiple linear regression (GA-MLR) is used to select an optimal subset of descriptors that have significant contribution to the overall UFL property from the large pool of calculated descriptors. The best resulted model is a four-variable multilinear model with a defined applicability range. The average absolute error and root-mean-square error obtained for the external test set are 1.75 vol % and 2.77, respectively. The proposed model can be used to predict the UFL of organic compounds with only four preselected theoretical descriptors which can be directly calculated from molecular structure alone.