Predicting normal densities of amines using quantitative structure-property relationship (QSPR)

This paper reports on a quantitative structure-property relationship (QSPR) approach applied to predict the normal density of amines. A heuristic method was applied to a dataset of experimental densities measured for more than 140 amines extracted from the literature. Statistical processing was performed to find the best correlations. The resulting model was validated successfully based on an external test set. The QSPR analysis showed the importance of intrinsic density, obtained by dividing molecular weight by calculated molecular volume. Further improvement of the model's predictive ability was achieved by introducing descriptors quantifying the influence of intermolecular volume. It has been shown that a simple correlation equation is sufficient to predict the density of amines with satisfactory accuracy. The equation requires only three descriptors: intrinsic density, overall or summation solute hydrogen bond acidity and a sum of intrinsic state values, each of which can be easily calculated using the Simplified Molecular Input Line Entry System Specification (SMILES). The equation provided may be applied to determine the density of amines or their derivatives which are unavailable or unknown.