A General Treatment of Solubility. 2. QSPR Prediction of Free Energies of Solvation of Specified Solutes in Ranges of Solvents

Abstract

As part of our general QSPR treatment of solubility (started in the preceding paper), we now present quantitative relationships between solvent structures and the solvation free energies of individual solutes. Solvation free energies of 80 diverse organic solutes are each modeled in a range from 15 to 82 solvents using our CODESSA PRO software. Significant correlations (in terms of squared correlation coefficient) are found for all the 80 solutes: the best fit is obtained for n-propylamine (R 2 ) 0.996); the lowest R 2 corresponds to toluene (0.604).

A General Treatment of Solubility. 3. Principal Component Analysis (PCA) of the Solubilities of Diverse Solutes in Diverse Solvents

“…A detailed description of the methodology for the development of the QSPRs has been given in our previous publications. 14,17 Principal component analysis (PCA) is one of the bestknown multivariate exploratory techniques extensively used in different areas of chemistry. [18][19][20] The PCA reveals internal relations between characteristics of a class of compounds (objects) and hence enables drastic reduction of the dimensionality of the original raw data.…”

“…The QSPR models for the solubility of single solutes in a range of solvents ("models for solvents") 14 and also the QSPR models for specified solutes in single solvents ("models for solutes") 17 using the general solubility matrix (HM0) have been further developed and revised in the present study. The newly developed QSPRs were used to predict the missing values of the logL for a so-called small…”

“…Significant revisions in the previously reported QSPR models 14,17 were made for the following reasons: (i) new experimental solubility data became available since our last QSPR modeling, and this lead to the extension of the general solubility matrix to 397 solutes × 154 solvents, as described above; (ii) there were mistakes in the 3D structural representation (used in ref 14) of nine of the compounds as detailed in the Supporting Information (SI-C); (iii) 16 experimental logL values were corrected in the general matrix (see Supporting Information SI-D); (iv) new features implemented into Codessa Pro were utilized: e.g. the BMLR algorithm [25][26][27] for the development of the QSPR models have been added and the pool of calculated descriptors has been increased with up to 40 hydrogen-bonding descriptors.…”

A General Treatment of Solubility 4. Description and Analysis of a PCA Model for Ostwald Solubility Coefficients

“…Therefore, considering the amount and diversity of the compounds in the external validation set, one can presume our previous predictions [7][8][9] are reliable. This demonstrates potential for the approach that encouraged analysis of the results and further exploration of the log L matrix.…”

“…The first two parts focused on the theoretical foundations, data gathering, and multilinear quantitative structure-property relationship (QSPR) modeling of a series of solvents 7 and solutes. 8 The third part utilized the derived QSPR models and provided a systematic approach to predict missing data points using a combination of QSPR and principal component analysis (PCA) methods. 9 Using this combined approach, several regions of the data matrix were filled by predictions from QSPR models.…”