Combining unsupervised and supervised artificial neural networks to predict aquatic toxicity

Giuseppina Gini, Marian Viorel Craciun, Christoph König, Emilio Benfenati

Research output: Contribution to journalArticle

31 Citations (Scopus)

Abstract

Most quantitative structure-activity relationship (QSAR) models are linear relationships and significant for only a limited domain of compounds. Here we propose a data-driven approach with a flexible combination of unsupervised and supervised neural networks able to predict the toxicity of a large set of different chemicals while still respecting the QSAR postulates. Since QSAR is applicable only to similar compounds, which have similar biological and physicochemical properties, large numbers of compounds are clustered before building local models, and local models are ensembled to obtain the final result. The approach has been used to develop models to predict the fish toxicity of Pimephales promelas and Tetrahymena pyriformis, a protozoan.

Original languageEnglish
Pages (from-to)1897-1902
Number of pages6
JournalJournal of Chemical Information and Computer Sciences
Volume44
Issue number6
DOIs
Publication statusPublished - Nov 2004

Fingerprint

Toxicity
Neural networks
Fish

ASJC Scopus subject areas

  • Chemistry(all)
  • Information Systems
  • Computer Science Applications
  • Computational Theory and Mathematics

Cite this

Combining unsupervised and supervised artificial neural networks to predict aquatic toxicity. / Gini, Giuseppina; Craciun, Marian Viorel; König, Christoph; Benfenati, Emilio.

In: Journal of Chemical Information and Computer Sciences, Vol. 44, No. 6, 11.2004, p. 1897-1902.

Research output: Contribution to journalArticle

Gini, Giuseppina ; Craciun, Marian Viorel ; König, Christoph ; Benfenati, Emilio. / Combining unsupervised and supervised artificial neural networks to predict aquatic toxicity. In: Journal of Chemical Information and Computer Sciences. 2004 ; Vol. 44, No. 6. pp. 1897-1902.
@article{69ff947bc35849a6ac0be347a97b01a0,
title = "Combining unsupervised and supervised artificial neural networks to predict aquatic toxicity",
abstract = "Most quantitative structure-activity relationship (QSAR) models are linear relationships and significant for only a limited domain of compounds. Here we propose a data-driven approach with a flexible combination of unsupervised and supervised neural networks able to predict the toxicity of a large set of different chemicals while still respecting the QSAR postulates. Since QSAR is applicable only to similar compounds, which have similar biological and physicochemical properties, large numbers of compounds are clustered before building local models, and local models are ensembled to obtain the final result. The approach has been used to develop models to predict the fish toxicity of Pimephales promelas and Tetrahymena pyriformis, a protozoan.",
author = "Giuseppina Gini and Craciun, {Marian Viorel} and Christoph K{\"o}nig and Emilio Benfenati",
year = "2004",
month = "11",
doi = "10.1021/ci0401219",
language = "English",
volume = "44",
pages = "1897--1902",
journal = "Journal of Chemical Information and Computer Sciences",
issn = "0095-2338",
publisher = "American Chemical Society",
number = "6",

}

TY - JOUR

T1 - Combining unsupervised and supervised artificial neural networks to predict aquatic toxicity

AU - Gini, Giuseppina

AU - Craciun, Marian Viorel

AU - König, Christoph

AU - Benfenati, Emilio

PY - 2004/11

Y1 - 2004/11

N2 - Most quantitative structure-activity relationship (QSAR) models are linear relationships and significant for only a limited domain of compounds. Here we propose a data-driven approach with a flexible combination of unsupervised and supervised neural networks able to predict the toxicity of a large set of different chemicals while still respecting the QSAR postulates. Since QSAR is applicable only to similar compounds, which have similar biological and physicochemical properties, large numbers of compounds are clustered before building local models, and local models are ensembled to obtain the final result. The approach has been used to develop models to predict the fish toxicity of Pimephales promelas and Tetrahymena pyriformis, a protozoan.

AB - Most quantitative structure-activity relationship (QSAR) models are linear relationships and significant for only a limited domain of compounds. Here we propose a data-driven approach with a flexible combination of unsupervised and supervised neural networks able to predict the toxicity of a large set of different chemicals while still respecting the QSAR postulates. Since QSAR is applicable only to similar compounds, which have similar biological and physicochemical properties, large numbers of compounds are clustered before building local models, and local models are ensembled to obtain the final result. The approach has been used to develop models to predict the fish toxicity of Pimephales promelas and Tetrahymena pyriformis, a protozoan.

UR - http://www.scopus.com/inward/record.url?scp=10044293032&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=10044293032&partnerID=8YFLogxK

U2 - 10.1021/ci0401219

DO - 10.1021/ci0401219

M3 - Article

VL - 44

SP - 1897

EP - 1902

JO - Journal of Chemical Information and Computer Sciences

JF - Journal of Chemical Information and Computer Sciences

SN - 0095-2338

IS - 6

ER -