Title: Robust hybrid wrapper/filter biomarker discovery from gene expression data based on generalised island model
Authors: Anouar Boucheham; Mohamed Batouche; Souham Meshoul
Addresses: Computer Science Department, College of NTIC, Constantine 2 University, 25000 Constantine, Algeria ' Computer Science Department, College of NTIC, Constantine 2 University, 25000 Constantine, Algeria ' Computer Science Department, College of NTIC, Constantine 2 University, 25000 Constantine, Algeria
Abstract: In the last few years, researchers have dedicated growing attention to biomarker identification given due to its extreme importance in genomics and personalised medicine. In this paper, biomarker discovery is handled by means of a cooperative parallel and distributed approach based on metaheuristics. More specifically, metaheuristics are employed according to a generalised island model architecture, and a hybrid wrapper/filter mechanism is applied within the selection process. In addition, a new initialisation strategy based on an ensemble of filters is proposed to bootstrap the metaheuristics. Experiments on 12 publicly available cancer datasets have shown that our approach competes with and even outperforms recent state-of-the-art methods in terms of the predictive performance and also provides robust selection through the different similarity measures. Biological interpretation of the selected signature reveals that the proposed method guarantees the selection of highly informative genes for cancer diagnosis.
Keywords: gene expression analysis; biomarker identification; biomarkers; tumours; tumour classification; data mining; wrappers; filters; hybrid feature selection; bioinformatics; SVM classifier; support vector machines; information gain; gene selection; parallel nature-inspired metaheuristics; genomics data; bootstrapping; cancer diagnosis.
DOI: 10.1504/IJCBDD.2015.072075
International Journal of Computational Biology and Drug Design, 2015 Vol.8 No.3, pp.251 - 274
Received: 06 Aug 2014
Accepted: 18 Oct 2014
Published online: 30 Sep 2015 *