Impact of Normalization and Filtering on Gene Expression Data Analysis
Author Information
Author(s): Beyene Joseph, Hu Pingzhao, Parkhomenko Elena, Tritchler David
Primary Institution: University of Toronto
Hypothesis
How do different normalization and filtering methods affect the analysis of gene expression data?
Conclusion
The study found that normalization and filtering algorithms significantly influence genetic analysis of gene expression data.
Supporting Evidence
- The normalization methods had a strong effect on the gene expression correlation structure.
- Different genes were identified as top-ranking depending on the normalization method used.
- The overlap in probe sets extracted using the three normalization approaches varied from 10% to 43%.
- Heritability estimates from data processed with RMA were larger than those from GCRMA and PLIER.
Takeaway
This study looked at how different ways of processing gene data can change the results we get, showing that the method we choose really matters.
Methodology
The study applied three normalization methods (RMA, GCRMA, PLIER) and a correlation-based filtering algorithm to analyze gene expression data from lymphoblastoid cell lines.
Limitations
The study did not investigate whether the gene expression measurements are normally distributed, which may affect the findings.
Participant Demographics
Lymphoblastoid cell lines from 14 CEPH Utah families.
Statistical Information
P-Value
3.7*10-5
Statistical Significance
p≤3.7*10-5
Want to read the original?
Access the complete publication on the publisher's website