Related Experiment Video
Updated: Jul 19, 2026
14:58
High-Throughput Transcriptome Analysis for Investigating Host-Pathogen Interactions
Published on: March 5, 2022
Enrichment analysis in high-throughput genomics - accounting for dependency in the NULL
David L Gold1, Kevin R Coombes, Jing Wang
1Department of Statistics, Texas A&M University, 3134 TAMU, College Statio, TX 77843-3143, USA. dlgold@tamu.edu
Briefings in Bioinformatics
|November 2, 2006
Summary
This study investigates enrichment analysis (EA) in bioinformatics, finding that the common assumption of independence between Gene Ontology (GO) classes is not detrimental. Our derived methods confirm conventional EA results, ensuring robust biological interpretation of high-throughput genomics data.
Area of Science:
- Bioinformatics
- Genomics
- Computational Biology
Background:
- High-throughput genomics generates vast data, requiring methods to extract biological insights like biomarkers and pathways.
- Genes with similar expression profiles may share underlying biological mechanisms relevant to disease.
- Enrichment Analysis (EA) is a popular bioinformatics method to identify biological themes in gene sets using Gene Ontology (GO) classes.
Purpose of the Study:
- To evaluate the impact of the independence assumption in conventional EA statistical testing.
- To develop and validate a method for EA that relaxes the independence assumption between GO classes.
- To determine if conventional EA results are robust despite potential dependencies between GO classes.
Main Methods:
- Derived the exact null distribution for GO class enrichment testing by relaxing the independence assumption.
- Utilized well-established statistical theory for the derivation.
- Applied the developed method to publicly available high-throughput genomics datasets.
Main Results:
- The derived statistical test accounts for potential dependencies between GO classes.
- Results from the new method closely align with conventional EA approaches that assume independence.
- Demonstrated that the independence assumption in EA is generally not detrimental to results.
Conclusions:
- Conventional enrichment analysis methods, despite assuming independence between Gene Ontology classes, yield robust results.
- The derived statistical framework provides a more formally correct approach to GO enrichment testing.
- Bioinformaticians can be confident in using standard EA tools for interpreting genomics data.
