Simple method for assessment of activities of thrombin inhibitors through their molecular structure parameters

Thrombi (blood clots) form in blood vessels in thromboembolic disorders, which are among the main reasons for death in the world. A novel approach is presented to predict thrombin inhibitory activities ((log(10³/K_i) (nM)) of some classes of non-peptidic thrombin inhibitors. The largest reported data set of log(10³/K_i) for 260 thrombin inhibitors are used to derive and test the new model where it can be easily applied through a computer code. The new model is derived and tested based on 201 and 59 experimental data, respectively, where its reliability is established by external and internal validations. The reliability of the novel correlation is compared with the complex 3D-QSAR method CoMSIA based on donor hydrogen bond, electrostatic interactions, steric occupancy, local hydrophobicity, and acceptor hydrogen bond fields. The values of correlation coefficient (R²), and root mean squared error (RMSE) for 138/34 data of training/test sets, where the predicted results of complex CoMSIA calculations were available, are 0.9173/0.6010 (R²), and 0.2503/0.4911 (RMSE) as well as 0.8753/0.3888 (R²), and 0.3287/0.6358 (RMSE) for the new and CoMSIA models, respectively. Further statistical parameters also confirm high reliability, precision, accuracy, and the goodness-of-fit of the simple model.