Modeling Properties with Higher-Level Molecular Connectivity Descriptors
Lionello Pogliani · Journal of Chemical Information and Computer Sciences · 1998
The internal formal similarity of different higher-level molecular connectivity descriptors, also known as molecular connectivity terms, which can be derived by a trial-and-error procedure from a medium-sized set of eight molecular connectivity indices or a subset of it has been thoroughly analyzed. These molecular connectivity terms are interesting descriptors of a wide range of properties of biochemical but also organic and inorganic compounds: amino acids, purines, pyrimidines, alkanes, and inorganic salts. The trial-and-error procedure sometimes uncovers “dead-end” terms, that is, terms which cannot be used in multilinear combinations as they give rise to a worse description, like, e.g., the single term used to model the crystal density of amino acids and a supra term (a third-level descriptor) for the description of the solubility of the heterogeneous class of amino acids, purines, and pyrimidines. The formal similarity of many terms used to describe the different properties simplifies a lot of the otherwise rather lengthy trial-and-error procedure used to discover new meaningful higher-level descriptors. Thus, formally similar terms are capable to model, e.g., the pH at the isoelectric point, pI, the specific rotations of d - and l -amino acids in aqueous solution, the water solubility of a heterogeneous class of amino acids plus purines and pyrimidines, for which even a fourth-level descriptor can be designed, the unfrozen water content of a heterogeneous class of amino acids and inorganic salts. Further, also the motor octane number of alkanes and five different properties of the DNA−RNA bases (U, T, A, G, and C) can be modeled with formally similar second level descriptors: first and second singlet excitation energies Δ E 1 and Δ E 2, the first and second oscillator strengths, f 1 and f 2 of the first singlet excitation energy, and the molar absorption coefficient ε 260 .