Prediction of n-octanol–water partition coefficient of platinum (IV) complexes using correlation weights of fragments of local symmetry

The octanol–water partition coefficient (logP) of platinum (IV) complexes is an essential indicator of the biological activity of coordination compounds in the aspect of potential application for drug design. The additive scheme of the logP simulation using a Simplified Molecular Input-Line Entry System (SMILES) was tested in a previous study. Here, it is proposed to take into account fragments of local symmetry (FLS) in SMILES. FLS are recognized as groups “xyx,” “xyyx,” and “xyzyx.” The CORAL software (www.insilico.eu/coral) generates optimal descriptors. The optimal descriptor is calculated using the so-called correlation weights for different SMILES fragments. Expanding the list of SMILES fragments by adding the aforementioned local symmetry markedly improves the predictive potential of models for the n-octanol–water partition coefficient of platinum (IV) coordination compounds (without FLS, the average determination coefficient for validation set is 0.87 ± 0.04, using FLS, the average becomes 0.91 ± 0.05). Thus, the involving fragments of local symmetry can improve the predictive potential of logP-models.