Development of quantitative structure-activity relationship models for vapor pressure estimation using computed molecular descriptors

Abstract

Vapor pressure is an important property which is an indicator of chemical volatility, along with transport, partitioning, fate and distribution of environmental pollutants. Various models have been developed for the prediction of vapor pressure of chemicals using physicochemical and calculated structural properties. We have used different classes of graph theoretic indices, e.g., topostructural indices, topochemical indices, geometrical (3D) indices and, quantum chemical descriptors, for the development of predictive models for vapor based on a structurally diverse set of 469 chemicals. Initially, a set of 379 molecular descriptors was calculated using the software POLLY, Triplet, Sybyl, MOPAC, and Molconn-Z. Comparatively, three linear regression methodologies were used to develop hierarchical QSAR (HiQSAR) models, namely ridge regression (RR), principal components regression (PCR), and partial least squares (PLS) regression. The results indicate that, in general, RR outperforms PCR and PLS, and that the easily calculated topological descriptors are sufficient for the prediction of vapor pressure based on this large, diverse set of chemicals.

Quantitative Structure−Activity Relationship Modeling of Juvenile Hormone Mimetic Compounds for Culex Pipiens Larvae, with a Discussion of Descriptor-Thinning Methods

“…In contrast to this, we modeled diverse sets of structures together using a diverse set of computable descriptors, deleting only 2 out of 304 observations as outliers, and still, all the important features responsible for JH activity were picked up in the final model, and more importantly, this is achieved after descriptor thinning, and this bears testimony to our philosophy of QSAR, “A diverse set of compounds can be modeled with a diverse set of descriptors”. This phenomenon was observed in several of our earlier studies on partition coffiecients, ,, vapor pressure, , mutagenicity, and the boiling point of organic compounds. In the QSAR modeling of blood/air 51,52 and tissue/air partition coefficients 40 of structurally and physicochemically diverse (hydrophobic and hydrophilic) compounds, we showed that (a) a collection of diverse descriptors is capable of giving good quality QSARs for progressively heterogeneous sets of halogenated compounds and (b) a set of diverse descriptors yielded good quality models for hydrophobic and hydrophilic compounds as well as the combined set of hydrophobic and hydrophilic chemicals, where physicochemically based algorithms failed in the case of the hydrophilic subset.…”

Quantitative Correlation of Physical and Chemical Properties with Chemical Structure: Utility for Prediction

“…Basak et al used the hierarchical quantitative structure−activity relationship (HiQSAR) approach for the prediction of vapor pressures based on structural descriptors using topostructural and topochemical parameters and an additional parameter (HB 1 ) related to intermolecular interactions for the prediction of VP measured at 25 °C for 476 diverse chemicals taken from the ASTER (assessment tools for the evaluation of risk) database and obtained a ten-parameter model with R 2 = 84.3% and s = 0.29. Three linear regression methodologiesridge regression (RR), principal component regression (PCR), and partial least-squares (PLS)were used to develop HiQSAR models for a VP data set of 469 chemicals based on topological descriptors . The results indicated that the RR outperforms PCR and PLS.…”

Application of Multivariate Adaptive Regression Splines (MARSplines) for Predicting Hansen Solubility Parameters Based on 1D and 2D Molecular Descriptors Computed from SMILES String

“…e quality of the chosen optimal d(25, 2) model is characterized by the fitting criteria: R 2 � 0.9470, (R adj ) 2 � 0.9378, LOF � 0.3680, K xx � 0.4341, RMSE tr � 0.4293, MAE tr � 0.3239, F � 103.3872, and N � 130, and fulfils the following internal validation criteria: (Q loo ) 2 � 0.8601, RMSE cv � 0.6973, and MAE cv � 0.4309 [71,72]. which appeared in the p(23, 3) model, IC0 and MLFER_S are quite simple to interpret in the context of polarity HSP since IC0 index expresses the diversity (heterogeneity) of atomic types [81], while MLFER_S is associated with the dipolarity/ polarizability features of molecules [57,82,83]. Also autocorrelation descriptors GATS1e, GATS2e, and MATS1v deserve for special attention.…”