Combining Gene Expression Data from Multiple Microarray Studies
Author Information
Author(s): Choi Hyungwon, Shen Ronglai, Chinnaiyan Arul M, Ghosh Debashis
Primary Institution: University of Michigan
Hypothesis
The profile for distinguishing metastatic from nonmetastatic tumors can be used to predict aggressive cancer prognosis.
Conclusion
The proposed methods for meta-analysis effectively combine genomic data from multiple studies to identify gene signatures that distinguish metastatic tumors from primary tumors.
Supporting Evidence
- The study identified gene signatures that can predict metastasis-free survival times.
- The methods were validated using data from a breast cancer study.
- The POE transformations improved the reliability of findings by reducing the influence of single study artifacts.
Takeaway
This study shows how scientists can combine data from different cancer studies to find common patterns that help identify aggressive tumors.
Methodology
The study used a probabilistic framework with mixture models and two estimation methods (MCMC and EM) to analyze gene expression data.
Potential Biases
The study may be subject to biases due to the variability in gene expression measurement across different platforms.
Limitations
The methodology assumes that transformed expression measurements are directly combinable across studies, which may not always be valid.
Participant Demographics
The study included samples from various organ sites, specifically liver, lung, and prostate cancers.
Statistical Information
P-Value
0.005
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website