Stability constants of complexes of Zn ²⁺, Cd ²⁺, and Hg ²⁺ with organic ligands: QSPR consensus modeling and design of new metal binders

QSPR modeling of the stability constant log K of the complexes of Zn2+, Cd2+ and Hg2+ with various 556 (Zn2+), 347 (Cd2+) and 76 (Hg2+) organic ligands in water for the M2+ + L = (M2+)L equilibrium at 298 K and an ionic strength 0.1 M was performed. Two machine-learning methods were used: Multiple Linear Regression Analysis (MLR) and Partial Robust M-regression Algorithm (PRM). The PRM method was realized for consensus modeling using substructural molecular fragments (SMF) as descriptors. Using different types of SMF, ensembles of individual predictive MLR and PRM models were prepared to build consensus models (CM). The root mean squared error of test set predictions of fivefold cross-validations is 1.8 and 1.9 (Zn2+), 1.9 and 2.2 (Cd2+), 2.7 and 2.8 (Hg2+) for the MLR and PRM approaches correspondingly. Experimental log K values vary in the range of 0.8–21.9 (Zn2+), 0.9–23.3 (Cd2+) and 1.6–29.7 (Hg2+). Extra validation of the models has been performed on a set of ligands recently reported in the literature. The QSPR models are sampled for the design of new binders of the Zn2+, Cd2+ Hg2+ cations.