Predicting the net heat of combustion of organic compounds from molecular structures based on ant colony optimization

A quantitative structure–property relationship (QSPR) model for prediction of standard net heat of combustion was developed from molecular structures. A diverse set of 1650 organic compounds were employed as the studied dataset, and a total of 1481 molecular descriptors were calculated for each compound. The novel variable selection method of ant colony optimization (ACO) algorithm coupled with the partial least square (PLS) was employed to select optimal subset of descriptors that have significant contribution to the overall property of standard net heat of combustion from the large pool of calculated descriptors. As a result, four molecular descriptors were screened out as the input parameters, and a four-variable multi-linear model was finally constructed using multi-linear regression (MLR) method. The resulted squared correlation coefficient R2 of the model was 0.995 for the training set of 1322 compounds, and 0.996 for the external test set of 328 compounds, respectively. The results showed that an accurate prediction model for the net heat of combustion could be obtained by using the ant colony optimization method. Moreover, this study can provide a new way for predicting the net heat of combustion of organic compounds for engineering based on only their molecular structures.

Research highlights► The optimization method of ant colony optimization (ACO) algorithm was employed in predicting the heat of combustion of organic compounds for variable selection for the first time. ► The proposed model could be simply used to predict the heat of combustion of pure components from only the knowledge of the molecular structures, without using any experimental parameter. ► The proposed model could be used to reveal the quantitative relation between the heat of combustion and molecular structures of organic compounds, and identify and provide some insight into what structural features are most related to the property of the heat of combustion of organic compounds. ► The prediction model was developed based on a robust dataset containing 1650 samples. These compounds cover extensive space of chemical varieties, and are representative and general in molecular structures. Thus the developed model was considered to be robust, effective and widely applied. Moreover, the prediction performance of this model is satisfying.