Use of the index of ideality of correlation to improve aquatic solubility model

Models for water solubility are built up using so-called correlation weights of fragments of simplified molecular input-line entry system (SMILES), 2D graph invariants, and hierarchy of rings present in molecules. The approach is based on Monte Carlo optimization of the molecular features. Two versions of the optimization were examined: (i) the traditional version; and (ii) the Index of Ideality of Correlation (IIC) version. The statistical quality of models obtained with use of the IIC is comparable with models built up with application of physico-chemical endpoints and quantum mechanics descriptors recently suggested in the literature.