Rank-1 Bicluster Classifier
- Nasimeh Asgarian, AICML
A DNA-microarray measures the gene expression levels of tens of thousands of genes under different experimental conditions (samples). These values describe the unique properties to each cell type. The samples may have come from different time points, different organs, diseased or healthy tissues, or different individuals. This technology benefits biological research greatly in understanding of biological processes. It also has very important applications in pharmaceutical and clinical research.
Our goal is to learn a microarray classifier that can distinguish members of various classes, based on their expression levels. Unfortunately, the large number of genes and the small number of samples make analyzing microarray data very challenging. In this thesis we propose a method for sample classification by first reducing the dimensionality of the data matrix, using bi-clusters. A bi-cluster is a subset of genes and a subset of samples that have similar patterns based on the gene expression values in the microarray data. We also propose a novel algorithm for finding bi-clusters from the microarray data, based on the best rank-1 matrix approximation. We demonstrate that our method works effectively by comparing its prediction accuracy with that of other classifiers, including one using another bicluster algorithm.
Citation
N. Asgarian. "Rank-1 Bicluster Classifier". MSc Thesis, University of Alberta, January 2007.| Keywords: | Microarray, Bi-cluster, Machine learning, bioinformatics |
| Category: | MSc Thesis |
BibTeX
@mastersthesis{Asgarian:07,
author = {Nasimeh Asgarian},
title = {Rank-1 Bicluster Classifier},
School = {University of Alberta},
year = 2007,
}Last Updated: July 21, 2009Submitted by Nelson Loyola
