Co-clustering and visualization of gene expression data and gene ontology terms for Saccharomyces cerevisiae using self-organizing maps

TY - JOUR

T1 - Co-clustering and visualization of gene expression data and gene ontology terms for Saccharomyces cerevisiae using self-organizing maps

AU - Brameier, Markus

AU - Wiuf, Carsten

PY - 2007/4/1

Y1 - 2007/4/1

N2 - We propose a novel co-clustering algorithm that is based on self-organizing maps (SOMs). The method is applied to group yeast (Saccharomyces cerevisiae) genes according to both expression profiles and Gene Ontology (GO) annotations. The combination of multiple databases is supposed to provide a better biological definition and separation of gene clusters. We compare different levels of genome-wide co-clustering by weighting the involved sources of information differently. Clustering quality is determined by both general and SOM-specific validation measures. Co-clustering relies on a sufficient correlation between the different datasets. We investigate in various experiments how much GO information is contained in the applied gene expression dataset and vice versa. The second major contribution is a visualization technique that applies the cluster structure of SOMs for a better biological interpretation of gene (expression) clusterings. Our GO term maps reveal functional neighborhoods between clusters forming biologically meaningful functional SOM regions. To cope with the high variety and specificity of GO terms, gene and cluster annotations are mapped to a reduced vocabulary of more general GO terms. In particular, this advances the ability of SOMs to act as gene function predictors.

AB - We propose a novel co-clustering algorithm that is based on self-organizing maps (SOMs). The method is applied to group yeast (Saccharomyces cerevisiae) genes according to both expression profiles and Gene Ontology (GO) annotations. The combination of multiple databases is supposed to provide a better biological definition and separation of gene clusters. We compare different levels of genome-wide co-clustering by weighting the involved sources of information differently. Clustering quality is determined by both general and SOM-specific validation measures. Co-clustering relies on a sufficient correlation between the different datasets. We investigate in various experiments how much GO information is contained in the applied gene expression dataset and vice versa. The second major contribution is a visualization technique that applies the cluster structure of SOMs for a better biological interpretation of gene (expression) clusterings. Our GO term maps reveal functional neighborhoods between clusters forming biologically meaningful functional SOM regions. To cope with the high variety and specificity of GO terms, gene and cluster annotations are mapped to a reduced vocabulary of more general GO terms. In particular, this advances the ability of SOMs to act as gene function predictors.

KW - Clustering validation

KW - Clustering visualization

KW - Co-clustering

KW - Gene expression data

KW - Gene function prediction

KW - Gene ontology

KW - Saccharomyces cerevisiae yeast

KW - Self-organizing maps

UR - http://www.scopus.com/inward/record.url?scp=33847663266&partnerID=8YFLogxK

U2 - 10.1016/j.jbi.2006.05.001

DO - 10.1016/j.jbi.2006.05.001

M3 - Journal article

C2 - 16824804

AN - SCOPUS:33847663266

SN - 1532-0464

VL - 40

SP - 160

EP - 173

JO - Journal of Biomedical Informatics

JF - Journal of Biomedical Informatics

IS - 2

ER -