Is conformation a fundamental descriptor in QSAR? A case for halogenated anesthetics

Abstract

SummaryAn intriguing question in 3D-QSAR lies on which conformation(s) to use when generating molecular descriptors (MD) for correlation with bioactivity values. This is not a simple task because the bioactive conformation in molecule data sets is usually unknown and, therefore, optimized structures in a receptor-free environment are often used to generate the MD´s. In this case, a wrong conformational choice can cause misinterpretation of the QSAR model. The present computational work reports the conformational analysis of the volatile anesthetic isoflurane (2-chloro-2-(difluoromethoxy)-1,1,1-trifluoroethane) in the gas phase and also in polar and nonpolar implicit and explicit solvents to show that stable minima (ruled by intramolecular interactions) do not necessarily coincide with the bioconformation (ruled by enzyme induced fit). Consequently, a QSAR model based on two-dimensional chemical structures was built and exhibited satisfactory modeling/prediction capability and interpretability, then suggesting that these 2D MD´s can be advantageous over some three-dimensional descriptors.

Optimization of a Deep-Learning Method Based on the Classification of Images Generated by Parameterized Deep Snap a Novel Molecular-Image-Input Technique for Quantitative Structure–Activity Relationship (QSAR) Analysis

“…Until today, support vector machine, random forest, and artificial NN were needed to select a reasonable combination of features (corresponding to chemical structure descriptors in QSAR analysis) manually when learning (feature selection techniques). In many cases, it is extremely difficult to find the optimal solutions, since myriad (Manallack et al, 2010; Talevi et al, 2012; Guimarães et al, 2016; Fang et al, 2017). Therefore, various approximation methods have been developed to obtain an optimal combination for an approximate solution (Yap et al, 2007; Kulkarni et al, 2009).…”

3D perspective into MIA‐QSAR: A case for anti‐HCV agents

“…However, it is possible to include 3D information into the 2D descriptors as proved by Estrada and colleagues (2001), but the implications of this inclusion remain unclear (Estrada, Molina, & Perdomo‐López, ). Therefore, there is still no evidence that the inclusion of spatial information into bidimensional descriptors improves the results obtained in 2D‐QSAR analysis (Guimarães, Duarte, Silla, & Freitas, ).…”