Gene set internal coherence in the context of functional profiling.

Area of Science:

• Genomics
• Bioinformatics
• Systems Biology

Background:

• Functional profiling methods are widely used in high-throughput experiments like microarray analysis.
• These methods rely on biological information to define gene modules (e.g., Gene Ontology - GO, KEGG pathways).
• An implicit assumption is that genes within these modules exhibit co-expression, which is not always true.

Purpose of the Study:

• To assess the internal co-expression correlation within popular functional modules (GO and KEGG pathways).
• To evaluate the coherence of functional modules in terms of gene co-regulation.
• To determine if incorporating gene-function correlation can improve the detection of functional modules in gene expression data.

Main Methods:

• Analysis of a large collection of microarray data.
• Calculation of a coherence index to measure functional module co-regulation.
• Application of a logistic regression model using gene-to-function correlation information.

Main Results:

• An unexpectedly low level of internal correlation was observed within many GO and KEGG modules.
• Only about 30% of GO modules and 57% of KEGG modules showed higher internal correlation than expected by chance.
• The study provides a coherence index for measuring functional module co-regulation.

Conclusions:

• Many popular functional categories exhibit lower co-expression than anticipated, hindering detection by conventional methods.
• The study provides the first exhaustive analysis of internal co-expression in popular functional categories.
• Using gene-to-function correlation information enhances the performance of functional profiling methods compared to conventional enrichment techniques.