Reliable Discovery of Molecular Signatures
Author Information
Author(s): Nilsson Roland, Björkegren Johan, Tegnér Jesper
Primary Institution: Linköping University and Karolinska Institutet
Hypothesis
Can a statistical framework be developed to control false discovery rates in molecular signature discovery?
Conclusion
The developed statistical framework enables reliable discovery of molecular signatures from genome-wide data.
Supporting Evidence
- The method was able to discover molecular signatures with 5% FDR in three cancer data sets.
- High predictive accuracy can occur despite high false discovery rates.
- Signatures can be unstable even when the false discovery rate is low.
Takeaway
This study shows how to find important gene markers for diseases while making sure we don't mistakenly think unimportant genes are important.
Methodology
The study used simulation studies and applied a novel hypothesis testing procedure to control false discovery rates.
Potential Biases
The reliance on existing data sets may introduce bias in the findings.
Limitations
The method may require very large sample sizes for reasonable performance.
Participant Demographics
The study analyzed publicly available cancer gene expression data sets.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website