QSAR Study for the Fish Toxicity of Benzene Derivatives

DUCHOWICZ, P. R.; MARRUGO HERNANDEZ, J. J.; ORTIZ, E. V.; CASTRO, E. A.; VIVAS REYES, R.

Journal of the Argentine Chemical Society

ASOCIACIÓN QUÝMICA ARGENTINA

Año: 2009

0365-0375

We searched Quantitative Structure-Toxicity models for predicting the fish toxicity against Poecilia reticulata elicited by a diverse set of 92 benzene derivatives. The simultaneous linear regression analyzes on 1176 constitutional, topological, geometrical, electronic, and lipophilic molecular descriptors derived from the software Dragon lead to a three-parameter relationship characterized with correlation coefficient of calibration of R=0.953, Leave-one-out Cross Validation of Rloo=0.947, and test set validation of Rval=0.889, and compares fairly well with a previously reported model based on extended topo-chemical atom (ETA) indices. Our developed QSAR involves a topological descriptor as the most relevant variable for the set of chemicals, a 3D-MoRSE and a Radial Distribution Function descriptor that show low inter-correlations.