Using In-Silico Deconfounding for Biomarker Discovery in Blood Samples
Author Information
Author(s): Dirk Repsilber, Sabine Kern, Anna Telaar, Gerhard Walzl, Gillian F. Black, Joachim Selbig, Shreemanta K. Parida, Stefan H.E. Kaufmann, Marc Jacobsen
Primary Institution: Research Institute for the Biology of Farm Animals
Hypothesis
Can a non-negative matrix decomposition algorithm effectively detect differentially expressed genes in heterogeneous tissue samples?
Conclusion
The proposed deconfounding algorithm can enhance the detection of biomarkers in heterogeneous tissues, even under noisy conditions.
Supporting Evidence
- The algorithm was validated using experimental data from sorted blood cells.
- Simulation studies showed improved detection of differential gene expression compared to classical methods.
- Results indicated that quantile normalization enhances the performance of the deconfounding algorithm.
Takeaway
This study shows a way to find important signals in mixed blood samples by using a special computer method that helps separate the different types of cells.
Methodology
The study used gene expression data from blood samples and applied a non-negative matrix decomposition algorithm to detect differentially expressed genes.
Potential Biases
Potential bias exists if gene expression is influenced by the proportions of cell types in the tissue.
Limitations
The study assumes independence of gene expression profiles from cell type proportions, which may not hold true for all genes.
Participant Demographics
Participants included tuberculosis patients and healthy controls, with a total of 80 individuals.
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