A complete thermodynamic description of H‐bonding for drug design is proposed. The approach is based on applicability of multiplicative principle to enthalpy (ΔH) and free energy (ΔG) of complex formation by means of Ed(Ea) and Cd(Ca) factors: Where Ed(Ea) and Cd(Ca) characterize the proton donor (acceptor) ability of compounds. The estimation of E and C values for 163 proton donors and 195 proton acceptors was done, making it possible to predict ΔH and ΔG for 31785 reactions. The comparison of these data with experimental ones for 936 reactions was carried out: The substituent effect on H‐bonding ability of compounds was considered. It was established that varying of substituents at active sites can alter its proton donor (acceptor) properties over a wide range of Ed(E3, cd(ca) with sufficient overlap between values of factors corresponding to different types of active sites. One could propose this approach to be applied to quantitative estimation of thermodynamic functions of H‐bonding in substrate‐receptor complexes; also for QSAR modelling, where E and C or its linear combination might be used as molecular descriptors characterizing quantitatively the relative H‐bonding ability of biologically active compounds. The second type of applications is used by the authors when modelling QSAR by means of the program package CLARAS (Classification and Recognition of Active Structures).

Discrimination of the acute toxicity of chemicals to guppy (Poecilia reticulata), fathead minnow (Pimephales promelas) and rainbow trout (Oncorhynchus mykiss) with nonpolar narcosis and other Modes Of Action (MOA) was successfully carried out by means of discrimination analysis and structural similarity analysis. Good QSARs for acute aquatic toxicity of chemicals with nonpolar narcosis MOA for the three fish species were constructed, based on molecular polarizability and H-bond acceptor factors calculated by the HYBOT program. These latter relationships are useful alternatives to QSARs based on the experimental parameter log K o/w because they yield good statistics and because they are amenable to physicochemical interpretation. In this case, the acute aquatic toxicity of chemicals with nonpolar MOA is presented as the result of two contributions (molecular polarizability, which is volume-related, and hydrogen bonding) with opposite action: molecular polarizability increases acute nonpolar narcosis toxicity and H-bond acceptor ability decreases it.

Detailed critical analysis of publications devoted to QSPR of aqueous solubility is presented in the review with discussion of four types of aqueous solubility (three different thermodynamic solubilities with unknown solute structure, intrinsic solubility, solubility in physiological media at pH=7.4 and kinetic solubility), variety of molecular descriptors (from topological to quantum chemical), traditional statistical and machine learning methods as well as original QSPR models.

This paper is devoted to the classification of chemicals with polar narcosis Mode of Action (MOA) to guppy, fathead minnow and rainbow trout, and to creation of QSARs for toxicity prediction of such chemicals on the basis of physicochemical descriptors calculated by the HYBOT program. More than 85% of all chemicals were correctly classified (polar narcosis/other MOA, polar narcosis/baseline toxicity (nonpolar narcosis), polar narcosis þ baseline toxicity/other MOA) on the basis of structural similarity. Regression equations with good statistical criteria using HYBOT physicochemical descriptors were obtained for the toxicity of chemicals with polar narcosis MOA.Comparison of those equations with the equations for baseline toxicity permitted estimation of the essential contribution of baseline toxicity to the toxicity of all chemicals with polar narcosis MOA. The ratio of polar and nonpolar contributions depends on hydrophobity. Such behavior can be related to the increased opportunity of a hydrophobic chemical to move in all directions in the hydrocarbon-like interior of the membrane.

32 Quantitative Structure-Property Relationship (QSPR) models were constructed for prediction of aqueous intrinsic solubility of liquid and crystalline chemicals. Data sets contained 1022 liquid and 2615 crystalline compounds. Multiple Linear Regression (MLR), Support Vector Machine (SVM) and Random Forest (RF) methods were used to construct global models, and k-nearest neighbour (kNN), Arithmetic Mean Property (AMP) and Local Regression Property (LoReP) were used to construct local models. A set of the best QSPR models was obtained: for liquid chemicals with RMSE (root mean square error) of prediction in the range 0.50-0.60 log unit; for crystalline chemicals 0.80-0.90 log unit. In the case of global models the large number of descriptors makes mechanistic interpretation difficult. The local models use only one or two descriptors, so that a medicinal chemist working with sets of structurally-related chemicals can readily estimate their solubility. However, construction of stable local models requires the presence of closely related neighbours for each chemical considered. It is probable that a consensus of global and local QSPR models will be the optimal approach for construction of stable predictive QSPR models with mechanistic interpretation.