New combined approach for prediction of stability constants of metal-ligand complexes using thermodynamic radii of metal ions and ensembles of regression models

A new combined approach on the base of radii thermodynamic (RT) of metal ions and ensembles of regression models (ERM) was applied to predict the stability constants (log K) of the ML complexes of 26 metal ions (M) with various organic ligands (L) in water. ERM were built on largest and structurally diverse data sets for 416 (Cd2+), 396 (Co2+), 883 (Cu2+), 612 (Ni2+), and 568 (Zn2+) complexes. RT and ERM have been used to predict log K of the complexes of Ca2+, Cd2+, Co2+, Cu2+, Fe2+, La3+, Mg2+, Mn2+, Ni2+, Pb2+, VO2+, Y3+, Zn2+, Ce3+, Pr3+, Nd3+, Sm3+, Eu3+, Gd3+, Tb3+, Dy3+, Ho3+, Er3+, Tm3+, Yb3+, and Lu3+. The approach performs well on the 21 external test sets: root mean square error varies from 1.1 to 1.6 for 16 metal ions, from 1.7 to 2.0 for Eu3+, Dy3+, and Mn2+, 2.4 and 2.8 for Gd2+ and VO2+, respectively. The accuracy of predictions of the stability constants using the combined approach is comparable with that of QSPR methods and DFT calculations. A predictor of the stability constants log K of the ML complexes of 27 metal ions is publically available on the server https://vpsolovev.ru/programs/ as a program StaCoRT under Windows operating systems.