Email updates

Keep up to date with the latest news and content from BMC Bioinformatics and BioMed Central.

Open Access Highly Accessed Open Badges Methodology article

A Bayesian variable selection procedure to rank overlapping gene sets

Axel Skarman1, Mohammad Shariati12, Luc Jans1, Li Jiang1 and Peter Sørensen1*

Author Affiliations

1 Department of Molecular Biology and Genetics, Aarhus University, Blichers Allé 20, PO Box 50, Aarhus, Tjele DK-8830, Denmark

2 Department of Animal Science, Ferdowsi University of Mashhad, Mashhad, 91775, Iran

For all author emails, please log on.

BMC Bioinformatics 2012, 13:73  doi:10.1186/1471-2105-13-73

Published: 3 May 2012

Additional files

Additional file 1:

Matrix where the rows represent the genes with their Entrez gene identifiers. The columns represent the Kyoto Encyclopedia of Genes and Genomes pathways. The elements can have the values zero or one. The value zero means that the gene is not in the pathway whereas the value one means that the gene is in the pathway.

Format: CSV Size: 1.2MB Download file

Open Data

Additional file 2:

Details of the Bayesian analysis.

Format: PDF Size: 118KB Download file

This file can be viewed with: Adobe Acrobat Reader

Open Data

Additional file 3:

Table showing the full list of gene sets, defined according to KEGG pathways, ranked according to the variable selection method. “Odds_ratio” means the odds ratio between the prior and posterior probability for the gene set of being included in the model. This is used as a Bayes factor to judge the significance. “Inf” indicates infinity, which is the evaluated Bayes Factor when the posterior probability is 1. “Explained variance” is the average variance explained by the t-statistic per gene in the gene set.

Format: CSV Size: 13KB Download file

Open Data

Additional file 4:

The Venn diagram is showing the overlap of genes between the pathways ‘Focal Adhesion’ and ‘Extracellular Matrix (ECM)-receptor Interaction’. ‘Focal adhesion’ was highly ranked by the ANOVA method but had a low posterior probability (0.013) of being included in the model when using the Bayesian method. A plausible reason for this is the high overlap with the pathway ‘Extracellular Matrix (ECM)-receptor Interaction’, which had a posterior probability 1 of being included in the model. These results indicate that ‘Focal Adhesion’ was ranked high in the ANOVA due to ‘guilt by association’ with the ‘ECM-receptor interaction’ pathway.

Format: PDF Size: 13KB Download file

This file can be viewed with: Adobe Acrobat Reader

Open Data