CORAL: Monte Carlo based global QSAR modelling of Bruton tyrosine kinase inhibitors using hybrid descriptors.

Global QSAR modelling was performed to predict the pIC values of 233 diverse heterocyclic compounds as BTK inhibitors with the Monte Carlo algorithm of CORAL software using the DCW hybrid descriptors extracted from SMILES notations of molecules. The dataset of 233 BTK inhibitors was randomly split into training, invisible training, calibration and validation sets. The index of ideality of correlation was also applied to build and judge the predictability of the QSAR models. Eight global QSAR models based on the hybrid optimal descriptor using two target functions, i.e. TF (W = 0) and TF (W = 0.2) have been constructed. The statistical parameters of QSAR models computed by TF are more reliable and robust and were used to predict the pIC values. The model constructed for split 4 via TF is regarded as the best model and the numerical values of , , and are equal to 0.7981, 0.7429, 0.7898 and 0.6784, respectively. By internal and external validation techniques, the predictability and reliability of the designed models have been assessed. The structural attributes responsible for the increase and decrease of pIC of BTK inhibitors were also identified.