Challenges in Microarray Class Discovery
Author Information
Author(s): Eva Freyhult, Mattias Landfors, Jenny Önskog, Torgeir R. Hvidsten, Patrik Rydén
Primary Institution: Umeå University
Hypothesis
How do normalization, gene selection, and clustering methods affect the performance of cluster analysis in microarray data?
Conclusion
The choice of cluster analysis methods significantly impacts the ability to correctly cluster individuals based on gene expression profiles.
Supporting Evidence
- Normalization, gene selection, and clustering method significantly impact performance.
- Gene selection methods using high standard deviation or principal component analysis are preferred.
- Hierarchical clustering using Ward's method and k-means clustering achieved the highest adjusted Rand index.
Takeaway
This study looks at different ways to group genes and samples based on their expression data, showing that the methods we choose can really change the results.
Methodology
The study evaluated 2780 cluster analysis methods on seven publicly available 2-channel microarray data sets, considering various aspects like normalization, gene selection, and clustering methods.
Limitations
The study focuses on 2-channel experiments with a common reference design, which may not apply to other platforms or experimental designs.
Statistical Information
P-Value
p<0.001
Statistical Significance
p<0.001
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website