Comparative Classical QSAR Modeling of Anti‐HIV Thiocarbamates

Abstract

The present Quantitative Structure -Activity Relationships (QSAR) study attempts to explore the structural and physicochemical requirements of O-[2-(phthalimido)ethyl]-substituted-phenylthiocarbamates for cytoprotection data from MT-4-based assays using a Linear Free Energy Related (LFER) model of Hansch. QSAR models have been developed using electronic (Hammett s), hydrophobicity (p) and steric (molar refractivity and STERIMOL L, B 1 , and B 5 ) parameters of phenyl ring substituents of the compounds along with appropriate dummy variables. Statistical techniques like stepwise regression, Multiple Linear Regression (MLR) with Factor Analysis (FA) as the data preprocessing step (FA-MLR), Partial Least Squares (PLS), Principal Component Regression (PCR), Multiple Linear Regression with Genetic Function Approximation (GFA-MLR) and Genetic Partial Least Squares (G/PLS) model were applied to identify the structural and physicochemical requirements for the cytoprotection activity. The generated equations were statistically validated using leave-one-out technique. The quality of equations obtained from these techniques were of acceptable statistical qualities (explained variance ranging from 57.7 to 75.9%, while predicted variance ranging from 53.2 to 68.5%), while the best model came from the PLS technique. The PLS equation with variables selected based on standardized regression coefficients show explained and predicted variances of 75.9 and 68.1%, respectively, and the G/PLS explained and predicted variances of 75.1 and 68.5%, respectively, of the cytoprotection activity data. Both stepwise regression and GFA derived models show high intercorrelation among predictor variables used in the equations, while the FA-MLR derived model has comparatively lower statistical quality. The best models were also subjected to leave-25%-out crossvalidation. The results of the present QSAR study are in agreement with the observations of the previously reported docking experiment of thiocarbamate derivatives.

On the interpretation and interpretability of quantitative structure–activity relationship models

“…This could be correlated to the log P and could have been investigated in detail using the PLS interpretation technique. Leonard and Roy [66] described a set of models built to model the anti-HIV activity of a set of thiocarbamates. Though they used linear regression and PLS to develop predictive models, the interpretations provided were derived directly from the signs and magnitudes of the linear regression coefficients.…”

Developing a variation of 3D‐QSAR/MD method in drug design

“…For decades, QSAR Models have been developed from 1D/2D traditional models (considering global molecular properties for modeling) to six‐dimensional forms (including structural patterns, nonbonded interaction and solvent models) 13 . Despite the robust and stable results of the traditional QSAR models 14–17 , they suffer from serious limitations such as considering only one or two‐dimensional descriptors and neglecting the spatial features of the drugs. Besides, traditional QSAR models are only applicable to a congeneric series of chemical compounds.…”

Optimal Piecewise Linear Regression Algorithm for QSAR Modelling

“…The first QSAR models were built for small series of similar compounds using only a few quantitative features and aimed to discover a transparent relationship, preferably linear, between molecular structure and biological activity . Although this approach is still employed to design new drugs,, most recent studies propose models that consist of hundreds or thousands of molecular descriptors calculated from the 2D or 3D representations of molecules and are often built with non‐linear algorithms such as Neural Networks, Support Vector Machines with Gaussian kernels and Random Forests . These techniques usually predict the biological activity of compounds with better accuracy than linear methods, but they are often described as “black box”, i. e. the relation between chemical features and biological activity can not be obtained directly from the outcome of the algorithm .…”