Prediction of Aqueous Solubility of Organic Compounds Based on a 3D Structure Representation

Two quantitative models for the prediction of aqueous solubility of 1293 organic compounds were developed by a Multilinear Regression (MLR) analysis and a Back-Propagation (BPG) neural network. The molecules were described by a set of 32 values of a Radial Distribution Function (RDF) code representing the 3D structure and eight additional descriptors. The 1293 compounds were divided into a training set of 797 compounds and a test set of 496 compounds based on a Kohonen self-organizing neural network map. The obtained models show a good predictive power: for the test set, a correlation coefficient of 0.96 and a standard deviation of 0.59 were achieved by the back-propagation neural network approach.