Chemometric modeling of core-electron binding energies

Abstract

Structures of 31 small molecules were modeled at HF 6-31 * level and 325 atom-type descriptors of different nature, such as electronegativity, polarizability, energy, charge, density, and steric descriptors, were calculated from molecular formula, optimized geometries, and literature data. The descriptors were employed in PLS (partial least squares) regression modeling of 59 X1s (59 X1s = 1 B1s, 27 C1s, 9 N1s, 14 O1s, and 8 F1s) core-electron binding energies (CEBEs) as unique set and also as elemental sets (C1s, N1s, O1s, and F1s). Parsimonius PLS models were obtained for all data sets (Q 2 > 0.79, R 2 > 0.84, SEV < 1.10 eV). Exploratory analyses of the PLS data sets have shown that CEBEs possess three-dimensional character which is determined by the element type of the ionized atom, electronegativity character of its chemical environment, and intramolecular stereolectronic effects.

Basic validation procedures for regression models in QSAR and QSPR studies: theory and application

“…4, 2009 Introduction Multivariate regression models in chemistry and other sciences quantitatively relate a response (dependent) variable y to a block of predictor variables X, in the form of a mathematical equation y = f(X), where the predictors can be determined experimentally or computationally. Among the best known of such quantitative-X-y relationships (QXYR) are quantitative structure-activity relationships (QSAR) [1][2][3] and quantitative structure-property relationships (QSPR), 4,5 in which y is a biological response (QSAR) or physical or chemical property (QSPR), and any of the predictors, designated as descriptors, may account for a microscopic (i.e., determined by molecular structure) or a macroscopic property. QSAR has become important in medicinal chemistry, pharmacy, toxicology and environmental science because it deals with bioactive substances such as drugs and toxicants.…”

“…Finally, the statistical reliability of the models is numerically and graphically tested [32][33][34] in various procedures called by the common name of model validation, [35][36][37][38] accompanied by other relevant verifications and model interpretation. [3][4][5]39 Even though the terms validation and to validate are frequent in chemometric articles, these words are rarely explained. 40 Among detailed definitions of validation in chemometric textbooks, of special interest is that discussed by Smilde et al, 32 who pointed out that validation includes theoretical appropriateness, computational correctness, statistical reliability and explanatory validity.…”

Interplay of thermochemistry and Structural Chemistry, the journal (volume 17, 2006) and the discipline

“…An early study suggested the difference of the dissociation energies of neutral molecules and isoelectronically related ions, where the dissociation processes were chosen to be also isoelectronic [293]. Kiralj and Takahata [294] theoretically modeled the structures of 31 small molecules and calculated atom-type descriptors which were subsequently employed for chemometric modeling of core-electron binding energies using a partial least squares regression model. This model showed that core-electron binding energies possesses three-dimensional character which is determined by the element type of the ionized atom, electronegativity character of its chemical environment, and intramolecular stereoelectronic effects.…”

The twentieth year in Structural Chemistry

“…The experimental binding energies of 1s core-electrons of 31 small molecules (altogether 59 core-electrons including B1s, C1s, N1s, O1s and F1s) were correlated with calculated and atom-type descriptors from the literature by Kiralj and Takahata [70]. The partial least squares regression analyses using electronegativity, polarizability, energy, charge, density, and steric descriptors revealed the three-dimensional character of the core-electron binding energies determined by the element type, electronegativity of its chemical environment, and intramolecular stereoelectronic effects.…”