Application of MOLMAP approach for QSAR modeling of various biological activities using substituent electronic descriptors

Abstract

When using quantum chemical descriptors in quantitative structure-activity relationship (QSAR) studies, there is always a challenge between accuracy of calculation and the complexity and time of computation. Very recently, we proposed the use of substituents electronic descriptors (SEDs) instead of the electronic properties of whole molecule as new and expedite source of electronic descriptors. For instance, SED parameters can be calculated with the highest degree of accuracy in very low computation time. In this article, we used SED parameters in QSAR modeling of six different biological data sets including (i) the dissociation constants for a set of substituted imidazolines, (ii) the pKa of imidazoles, (iii) inverse agonist activity of indoles, (iv) the influenza virus inhibition activities of benzimidazoles, (v) inhibition of alcohol dehydrogenase by amides, and (vi) the natriuretic activity of sulfonamide. For poly-substituted molecules, SED parameters produce a vector of electronic descriptors for each substituent, and thus a matrix of SED parameters is obtained for each molecule. Consequently, a three-dimensional (3D) array is obtained by staking the descriptor data matrices of molecules beside each others. In addition to simple unfolding of the SED parameters, molecular maps of atom-level properties (MOLMAP) approach, as a novel data analysis method, was also applied to transfer 3D array of SED into new two-dimensional parameters using Kohonen network, following by genetic algorithm-based partial least square (GA-PLS) to connect a quantitative relationship between the Kohonen scores and biological activity. Accurate QSAR models were obtained by both approaches. However, the superiority of three-way analysis of SED parameters based on MOLMAP approach with respect to simple unfolding was obtained.