Development of an Automated SMILES Pattern Matching Program To Facilitate the Prediction of Thermophysical Properties by Group Contribution Methods

Group-contribution methods produce estimates of thermophysical properties from correlations that include summative contributions for the functional groups that make up a molecule. Accurately dividing or parsing a molecule into its constituent functional groups, as intended by the authors of group-contribution methods, can be tedious and error prone. An automated parsing algorithm, which accepts SMILES formulas of compounds and accurately parses these into constituent functional groups, has been developed to facilitate the use, comparison, and development of group-contribution methods. The algorithm is described with particular attention to the difficulties inherent in parsing large, multiring compounds. The methodology will be useful to others who are developing, testing, and using prediction techniques.