Utility and Limitations of Using Gene Expression Data to Identify Functional Associations

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Gene co-expression is a common method for inferring gene function via guilt-by-association.
• The reliability and applicability of co-expression for diverse biological processes and datasets remain unclear.

Purpose of the Study:

• To assess the utility and limitations of gene co-expression for recovering functional gene associations.
• To investigate how dataset characteristics influence functional inference from co-expression data.

Main Methods:

• Analyzed gene pair expression correlation within metabolic pathways in Arabidopsis thaliana.
• Evaluated the impact of dataset choice, annotation quality, gene function, similarity measures, and clustering algorithms.
• Validated co-expression cluster memberships using independent phenomics data.

Main Results:

• Many genes within the same metabolic pathway exhibit dissimilar expression profiles.
• Dataset type, annotation quality, and analytical methods significantly affect the recovery of functional associations.
• Larger datasets are not consistently more informative; dataset selection is critical.

Conclusions:

• Gene co-expression is a valuable tool but requires careful optimization of multiple parameters for robust functional association recovery.
• Exploring diverse dataset combinations and analytical approaches is essential for maximizing information from gene expression data.
• Co-expression clusters show biological relevance, as demonstrated by validation with phenomics data and leucine metabolism genes.