Molecular Fragments Variable Connectivity Index And Its Application In QSAR Study

Quantitative Structure Activity Relationship (QSAR) is an effective predictive method in the research and analysis between the characteristics of the basic structure of chemicals and their properties, such as physico-chemical properties, toxicity and biological activities. This method has been widely used and rapidly developed in explanation mechanism and prediction properties of unknown compounds. With the fast development of society and science, QSAR becomes a cutting-edge research in chemistry, environment, and life sciences and so on.Molecular descriptors used in QSAR mainly contain physico-chemical descriptors, quantum chemical descriptors and topological descriptors. Among these molecular descriptors, topological indices remain very promising in QSAR study as they are easily obtained, especially in cases where their usage significantly reduces the computation time and usually give good correlations to the modeled property. However, a problem in the previous study of topological descriptors is failure to differentiate the different kinds of atoms. This problem was solved with the introduction of variable topological indices which was proved to be more suitable for the modeled property. The aim of this study is to improve the variable connectivity index.(1) The atoms or functional groups in different characteristics of different chemical bonding play a dominant role in chemical activity of compounds. In order to distinguish their different roles in same properties, the molecular fragments are defined as the atoms or functional groups having different characteristics due to different chemical bonding, which were considered as the molecular structure unit for the major influence of property. The molecular fragment is regarded as the vertex of the topological diagram and endue with variable w_i as substitute for the topological matrix diagonal. Hence, molecular fragments variable connectivity index (mfVCI, X^mf), a new molecular descriptor, has been proposed based on molecular fragments.(2) The QSAR models of boiling points of 100 alcohols, rate constants for reactions of hydroxyl radicals of 39 organic compounds, toxicity (Vibrio fischeri pT₅₀) of 138 halogenated substituted-benzenes and 92 aromatic compounds have beenmodeled by ¹X^mf,¹X^f,^`X,¹X^v,Zagreb index, k_n^α,φand J, respectively. Leave one out,leave many out and external validations, founded by ¹X^mf , were applied to test the goodness of fit and predictive ability of QSAR models. The results showed that models developed from ¹X^mf have higher predictivity, robustness and explanation ability. Models founded with ¹X^mf are superior to other models through comparison of the residuals.