A compressed sensing based two-stage method for detecting epistatic interactions. (2016)

• Record Type: Journal Article
• Authors: Li, Shengjun; Shang, Junliang; Chen, Qinliang; Sun, Yan; Liu, Jin-Xing
• Is Part Of: International journal of data mining and bioinformatics. Volume 14:Number 4(2016)
• Page Start: 354
• Page End: 372
• Publication Date: 2016
• Subjects: epistatic interactions -- SNPs -- single nucleotide polymorphisms -- compressed sensing -- sparse representation -- mutual information -- bioinformatics -- epistatic interaction detection; Data mining -- Periodicals; Bioinformatics -- Periodicals; 006.312
• Languages: English
• ISSNs: 1748-5673

Abstract

Epistatic interactions of single nucleotide polymorphisms (SNPs) are believed to be important in revealing missing heritability of complex diseases. Detection of them is of great challenge since it is a high-dimensional and small-sample-size problem. In this paper, we propose a compressed sensing (CS) based two-stage method CSMiner for detecting epistatic interactions. It consists of two stages: screening stage and detecting stage. In screening stage, SNP selection is equivalent to CS reconstruction by considering SNP data and class labels as sensing matrix and measurement vector, respectively. Here, top ranking SNPs with high signal weights are retained. In detecting stage, mutual information is employed to exhaustively search epistatic interactions within the retained SNPs. Experiments of CSMiner are performed on both simulation data sets and a real age-related macular degeneration data set. Results demonstrate that CSMiner is effective and efficient in detecting epistatic interactions, and might be an alternative to existing methods.