Prediction of small-molecule compound solubility in organic solvents by machine learning algorithms

Rapid solvent selection is of great significance in chemistry. However, solubility prediction remains a crucial challenge. This study aimed to develop machine learning models that can accurately predict compound solubility in organic solvents. A dataset containing 5081 experimental temperature and solubility data of compounds in organic solvents was extracted and standardized. Molecular fingerprints were selected to characterize structural features. lightGBM was compared with deep learning and traditional machine learning (PLS, Ridge regression, kNN, DT, ET, RF, SVM) to develop models for predicting solubility in organic solvents at different temperatures. Compared to other models, lightGBM exhibited significantly better overall generalization (logS ± 0.20). For unseen solutes, our model gave a prediction accuracy (logS ± 0.59) close to the expected noise level of experimental solubility data. lightGBM revealed the physicochemical relationship between solubility and structural features. Our method enables rapid solvent screening in chemistry and may be applied to solubility prediction in other solvents.

Isavuconazole: Thermodynamic Evaluation of Processes Sublimation, Dissolution and Partition in Pharmaceutically Relevant Media

A temperature dependence of saturated vapor pressure of isavuconazole (IVZ), an antimycotic drug, was found by using the method of inert gas-carrier transfer and the thermodynamic functions of sublimation were calculated at a temperature of 298.15 K. The value of the compound standard molar enthalpy of sublimation was found to be 138.1 ± 0.5 kJ·mol−1. The IVZ thermophysical properties—melting point and enthalpy—equaled 302.7 K and 29.9 kJ mol−1, respectively. The isothermal saturation method was used to determine the drug solubility in seven pharmaceutically relevant solvents within the temperature range from 293.15 to 313.15 K. The IVZ solubility in the studied solvents increased in the following order: buffer pH 7.4, buffer pH 2.0, buffer pH 1.2, hexane, 1-octanol, 1-propanol, ethanol. Depending on the solvent chemical nature, the compound solubility varied from 6.7 × 10−6 to 0.3 mol·L−1. The Hansen s approach was used for evaluating and analyzing the solubility data of drug. The results show that this model well-described intermolecular interactions in the solutions studied. It was established that in comparison with the van’t Hoff model, the modified Apelblat one ensured the best correlation with the experimental solubility data of the studied drug. The activity coefficients at infinite dilution and dissolution excess thermodynamic functions of IVZ were calculated in each of the solvents. Temperature dependences of the compound partition coefficients were obtained in a binary 1-octanol/buffer pH 7.4 system and the transfer thermodynamic functions were calculated. The drug distribution from the aqueous solution to the organic medium was found to be spontaneous and entropy-driven.

Solubilization and thermodynamic properties of simvastatin in various micellar solutions of different non-ionic surfactants: Computational modeling and solubilization capacity

The aim of this work was to solubilize simvastatin (SIM) using different micellar solutions of various non-ionic surfactants such as Tween-80 (T80), Tween-20 (T20), Myrj-52 (M52), Myrj-59 (M59), Brij-35 (B35) and Brij-58 (B58). The solubility of SIM in water (H2O) and different micellar concentrations of T80, T20, M52, M59, B35 and B58 was determined at temperatures T = 300.2 K to 320.2 K under atmospheric pressure p = 0.1 MPa using saturation shake flask method. The experimental solubility data of SIM was regressed using van’t Hoff and Apelblat models. The solubility of SIM (mole fraction) was recorded highest in M59 (1.54 x 10−2) followed by M52 (6.56 x 10−3), B58 (5.52 x 10−3), B35 (3.97 x 10−3), T80 (1.68 x 10−3), T20 (1.16 x 10−3) [the concentration of surfactants was 20 mM in H2O in all cases] and H2O (1.94 x 10−6) at T = 320.2 K. The same results were also recorded at each temperature and each micellar concentration of T80, T20, M52, M59, B35 and B58. “Apparent thermodynamic analysis” showed endothermic and entropy-driven dissolution/solubilization of SIM in H2O and various micellar solutions of T80, T20, M52, M59, B35 and B58.

Melting properties of peptides and their solubility in water. Part 1: dipeptides based on glycine or alanine

Combination of the PC-SAFT and FSC methods allows for accurate prediction of dipeptide solubility in water in a wide temperature range without the need to fit any model parameters to experimental solubility data.