Assessing Kernel Density Clustering for Gene Expression Data
Author Information
Author(s): Guoping Shu, Beiyan Zeng, Yiping P. Chen, Oscar H. Smith
Primary Institution: Reid Research Centre, Pioneer Hi-Bred International Inc.
Hypothesis
Can kernel density clustering effectively analyze gene expression profile data compared to traditional methods?
Conclusion
Kernel density clustering outperforms traditional clustering methods in recovering clusters from both simulated and real gene expression data.
Supporting Evidence
- The kernel density clustering method showed excellent performance in recovering clusters from simulated data.
- It was the most robust method for analyzing noisy expression profile data compared to other methods.
- Kernel density clustering effectively grouped large real expression profile data sets into compact and well-isolated clusters.
- The method does not require the user to specify the number of clusters beforehand.
Takeaway
This study shows a new way to group genes based on their expression patterns, which works better than older methods, especially when the data is noisy.
Methodology
The study compared kernel density clustering with hierarchical clustering, K-means clustering, and multivariate mixture model-based clustering using simulated and real gene expression data.
Limitations
The method may be less accurate for very small clusters due to the need for multiple data points for density estimation.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website