Premium
Consensus QSAR Related to Global or MOA Models: Application to Acute Toxicity for Fish
Author(s) -
Lozano Sylvain,
HalmLemeille MariePierre,
Lepailleur Alban,
Rault Sylvain,
Bureau Ronan
Publication year - 2010
Publication title -
molecular informatics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.481
H-Index - 68
eISSN - 1868-1751
pISSN - 1868-1743
DOI - 10.1002/minf.201000104
Subject(s) - quantitative structure–activity relationship , in silico , identification (biology) , fish <actinopterygii> , computer science , computational biology , bayesian probability , machine learning , artificial intelligence , data mining , biochemical engineering , biology , ecology , engineering , gene , genetics , fishery
Under REACH legislation, alternative methods (in silico or in vitro) like QSAR (Quantitative Structure‐Activity Relationships) models are expected to play a significant role. QSARs are based on the assumption that substances with similar chemical structures may have the same biological activities. However, identification of chemical classes could be problematic because chemicals often exhibit different chemical moieties, thereby confounding efforts to achieve a meaningful classification. This publication is focus on the notion of global model with the integration of a recent genetic algorithm for the generation of QSAR models. Starting from three datasets (EPAFHM, ECBHPV, AQUIRE), prediction of acute toxicity for fish ( Pimephales promelas ) with a global consensus model was carried out leading to very interesting statistics. The integration of the notion of Mode of Action was the second point of this study. A Bayesian classification associated to the genetic algorithm for consensus models was created leading to a good estimation of toxicity associated to derivatives with nonspecific MOA.