Heading down the wrong pathway: on the influence of correlation within gene sets

Area of Science:

• Bioinformatics
• Genomics
• Statistical Genetics

Background:

• Microarray analysis commonly tests for gene set overrepresentation among significant genes.
• Existing methods often assume gene independence, which is violated by known gene correlations.

Purpose of the Study:

• To demonstrate the impact of gene correlation on microarray gene set testing.
• To evaluate the performance of independence-based versus resampling-based methods.

Main Methods:

• Meta-analysis of over 200 Gene Expression Omnibus datasets.
• Comparison of independence assumption-based gene set testing with array resampling approach.

Main Results:

• Independence assumption-based methods yield high false positive rates, especially with correlated gene sets.
• Array resampling properly controls false positive rates, yielding more reliable gene set findings.
• High internal gene correlation increases the likelihood of false positives in standard methods.

Conclusions:

• Current gene set testing results in the literature may be unreliable due to violated independence assumptions.
• Adoption of resampling-based gene set testing is recommended for the biomedical literature.
• Resampling methods enhance confidence and facilitate pathway-based interpretation of microarray data.