Related Experiment Videos
A mathematical and computational framework for quantitative comparison and integration of large-scale gene expression
Christopher E Hart1, Lucas Sharenbroich, Benjamin J Bornstein
1Division of Biology, California Institute of Technology, Pasadena, CA 91125, USA.
Nucleic Acids Research
|May 12, 2005
Summary
Different gene clustering algorithms yield varied results. This study introduces CompClust, a framework to quantify, compare, and visualize these clusterings, integrating them with other data for biological insights.
Area of Science:
- Bioinformatics
- Computational Biology
- Systems Biology
Background:
- Gene expression analysis often starts with clustering, but different algorithms produce substantially different results.
- The variability in gene clustering raises questions about the reliability and biological significance of findings.
- A quantitative method is needed to compare clusterings and relate them to biological data.
Purpose of the Study:
- To develop a mathematical and computational framework for quantifying, comparing, visualizing, and mining gene clusterings.
- To assess the quality and overlap of different clustering results.
- To integrate gene expression clustering with other large-scale data like DNA motifs and protein-DNA interactions.
Main Methods:
- Coupling confusion matrices with linear assignment and normalized mutual information scores to quantify clustering differences.
- Utilizing receiver operator characteristic analysis for cluster quality and overlap assessment.
- Developing the CompClust software tool (version 1.0) with a library of clustering algorithms.
Main Results:
- The CompClust framework effectively quantifies and maps differences between gene clusterings.
- Receiver operator characteristic analysis proved useful for evaluating cluster quality and overlap.
- CompClust successfully integrated gene expression patterns with DNA motif and protein-DNA interaction data.
- Analysis of yeast cell cycle data revealed underlying data structures and identified G1 regulatory modules.
Conclusions:
- The developed framework and CompClust software provide a systematic approach to address the challenges of gene clustering variability.
- These tools enable a more robust interpretation of gene expression data by comparing clusterings and integrating diverse biological datasets.
- The integration of expression data with motif and interaction data facilitates the discovery of functional gene regulatory modules.