SGC method for predicting the standard enthalpy of formation of pure compounds from their molecular structures

• ΔH°f is predicted from the molecular structure of the compounds alone.
• ANN-SGC model predicts ΔH°f with a correlation coefficient of 0.99.
• ANN-MNLR model predicts ΔH°f with a correlation coefficient of 0.90.
• Better definition of the atom-type molecular groups is presented.
• The method is better than others in terms of combined simplicity, accuracy and generality.

A theoretical method for predicting the standard enthalpy of formation of pure compounds from various chemical families is presented. Back propagation artificial neural networks were used to investigate several structural group contribution (SGC) methods available in literature. The networks were used to probe the structural groups that have significant contribution to the overall enthalpy of formation property of pure compounds and arrive at the set of groups that can best represent the enthalpy of formation for about 584 substances. The 51 atom-type structural groups listed provide better definitions of group contributions than others in the literature. The proposed method can predict the standard enthalpy of formation of pure compounds with an AAD of 11.38 kJ/mol and a correlation coefficient of 0.9934 from only their molecular structure. The results are further compared with those of the traditional SGC method based on MNLR as well as other methods in the literature.