Premium
Topological characteristics of iterated line graphs in QSAR problem: Octane numbers of saturated hydrocarbons
Author(s) -
Zakharov Anton B.,
Dyachenko Anastasia V.,
Ivanov Vladimir V.
Publication year - 2019
Publication title -
journal of chemometrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.47
H-Index - 92
eISSN - 1099-128X
pISSN - 0886-9383
DOI - 10.1002/cem.3169
Subject(s) - quantitative structure–activity relationship , iterated function , topological index , vertex (graph theory) , mathematics , octane , line graph , graph , computer science , discrete mathematics , combinatorics , machine learning , chemistry , mathematical analysis , organic chemistry
Capabilities of line (edge) graph in quantitative structure‐activity relationship (QSAR) problem for constructing predictive models are demonstrated. The basic idea for proposed approach is to describe molecular structure using generated graph sequence: vertex and chain of line graphs. Every following graph in the sequence corresponds to the structure where vertices appear to be edges of the previous one. For every graph, it is proposed to compute corresponding topological and informational descriptors. Employing latter for derivation of predictive equations allows to realize progressive hierarchy of models. As a test problem, the forecast of research octane numbers for saturated hydrocarbons is chosen. Among the descriptors selected to be used for building regression models well‐known Randic connectivity index ( χ (1) ), first Zagreb index ( ZM1 ) and theoretic‐informational index of vertex complexity ( Vc ) are employed. Derived hierarchy of regression equations shows systematic improvement in the description of desired property as every new graph is included in consideration. Determination coefficients for training sample and computed via leave‐one‐out cross‐validation procedure reveal that four‐parametric equation with Vc descriptor has satisfactory forecasting facilities. Efficiency of proposed approach was also demonstrated on the example of artificial neural networks.