@inproceedings{e14677c6bfd94b718b264d03bb62aff2,
title = "A wrapper-based feature selection method for ADMET prediction using evolutionary computing",
abstract = "Wrapper methods look for the selection of a subset of features or variables in a data set, in such a way that these features are the most relevant for predicting a target value. In chemoinformatics context, the determination of the most significant set of descriptors is of great importance due to their contribution for improving ADMET prediction models. In this paper, a comprehensive analysis of descriptor selection aimed to physicochemical property prediction is presented. In addition, we propose an evolutionary approach where different fitness functions are compared. The comparison consists in establishing which method selects the subset of descriptors that best predicts a given property, as well as maintaining the cardinality of the subset to a minimum. The performance of the proposal was assessed for predicting hydrophobicity, using an ensemble of neural networks for the prediction task. The results showed that the evolutionary approach using a non linear fitness function constitutes a novel and a promising technique for this bioinformatic application.",
keywords = "Feature selection, Genetic algorithms, Hydrophobicity, QSAR",
author = "Soto, {Axel J.} and Cecchini, {Roc{\'i}o L.} and Vazquez, {Gustavo E.} and Ignacio Ponzoni",
year = "2008",
doi = "10.1007/978-3-540-78757-0_17",
language = "Ingl{\'e}s",
isbn = "3540787569",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "188--199",
booktitle = "Evolutionary Computation, Machine Learning and Data Mining in Bioinformatics - 6th European Conference, EvoBIO 2008, Proceedings",
note = "6th European Conference on Evolutionary Computation, Machine Learning, and Data Mining in Bioinformatics, EvoBIO 2008 ; Conference date: 26-03-2008 Through 28-03-2008",
}