Related Experiment Videos
GONOME: measuring correlations between GO terms and genomic positions
Stefan M Stanley1, Timothy L Bailey, John S Mattick
1Institute for Molecular Bioscience, University of Queensland, Brisbane 4072, Australia. stanley@imb.uq.edu.au
BMC Bioinformatics
|March 1, 2006
Summary
Gene Ontology (GO) term analysis is improved by the GONOME algorithm, which accounts for gene length to accurately identify genomic position associations. This method corrects biases found in previous approaches, ensuring more reliable biological insights.
Area of Science:
- Genomics
- Bioinformatics
- Computational Biology
Background:
- Existing methods for identifying over-represented Gene Ontology (GO) terms in gene sets often treat genes as equally probable, similar to "balls in a bag" models.
- This "balls in a bag" approach, suitable for microarray data, is inadequate for analyzing genomic positions due to variations in gene and intergenic region lengths.
- Such methods can lead to inaccurate conclusions regarding the correlation between GO terms and genomic locations.
Purpose of the Study:
- To introduce GONOME, a novel algorithm designed to accurately determine GO terms significantly associated with genomic positions.
- To address the limitations of existing methods by incorporating gene length into the analysis of genomic position-GO term associations.
- To demonstrate the potential of GONOME in discovering novel genomic features and correcting erroneous biological interpretations.
Main Methods:
- Development of the GONOME algorithm, which analyzes genomes annotated with gene start and end positions.
- Comparison of GO term enrichment results with and without considering gene lengths using simulated and real genomic data.
- Application of an extended GONOME approach for whole-genome motif discovery.
Main Results:
- GONOME accurately identifies significantly associated GO terms by considering gene lengths, mitigating biases present in length-agnostic methods.
- Analysis revealed that previously reported significant associations between GO terms (e.g., "development") and human CpG islands disappear when gene length is accounted for.
- The algorithm successfully identified appropriate GO terms for the proteasome-associated control element (PACE) upstream activating sequence in S. cerevisiae and discovered novel motifs linked to translation.
Conclusions:
- GONOME provides a robust method for extracting over-represented GO terms from genomic positions, explicitly accounting for gene size to prevent bias.
- The algorithm corrects erroneous conclusions stemming from the inappropriate use of existing methods that ignore gene length.
- GONOME can be utilized to identify novel genomic features significantly associated with specific gene categories.