Improving Gene Expression Data Normalization with Kernel Density Methods
Author Information
Author(s): Hsieh Wen-Ping, Chu Tzu-Ming, Lin Yu-Min, Wolfinger Russell D
Primary Institution: Institute of Statistics, National Tsing Hua University
Hypothesis
Can kernel density weighted loess normalization improve the performance of detection within asymmetrical gene expression data?
Conclusion
Methods based on invariant sets are better able to resolve the problem of asymmetry in gene expression data.
Supporting Evidence
- KDL and KDQ methods provided the best performance among all approaches tested.
- The proposed methods improved the detection power of differentially expressed genes.
- Normalization methods based on invariant sets showed superiority over traditional methods.
Takeaway
This study shows that new methods can help scientists better understand gene expression by correcting for biases in the data.
Methodology
The study proposes two novel approaches, KDL and KDQ, based on kernel density estimation to improve normalization of gene expression data.
Potential Biases
The presence of empty genes and the potential for global shifts in gene ranking can introduce bias.
Limitations
The methods may require careful tuning of parameters and may not perform well if the assumptions about gene distribution are violated.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website