Related Experiment Videos
Interactively optimizing signal-to-noise ratios in expression profiling: project-specific algorithm selection and
Jinwook Seo1, Marina Bakay, Yi-Wen Chen
1Research Center for Genetic Medicine, Children's National Medical Center, MD, USA.
Bioinformatics (Oxford, England)
|May 1, 2004
Summary
Choosing the right mRNA profiling algorithm is crucial for accurate gene expression analysis. This study shows that optimizing probe set algorithms based on project-specific noise levels, using Microarray Suite (MAS) 5.0 detection p-values, significantly improves data analysis performance.
Area of Science:
- Bioinformatics
- Computational Biology
- Genomics
Background:
- Affymetrix microarrays use probe sets for mRNA profiling, with various algorithms interpreting these sets to derive expression signals.
- Algorithm performance varies in accuracy and sensitivity, often optimized using limited control data.
- Project-specific confounding noise may necessitate different algorithm choices for optimal results.
Purpose of the Study:
- To investigate if confounding noise levels in mRNA profiling projects influence the optimal choice of probe set algorithm.
- To evaluate the impact of using Microarray Suite (MAS) 5.0 probe set detection p-values as a weighting function to improve algorithm performance.
- To develop an interactive tool for optimizing Affymetrix analysis parameters and signal-to-noise ratios.
Main Methods:
- Developed the Hierarchical Clustering Explorer 2.0 (HCE2) software for interactive analysis of Affymetrix data.
- Tested five probe set algorithms with and without MAS 5.0 p-value weighting on two distinct datasets: human muscle (high noise) and mouse lung (low noise).
- Assessed algorithm performance using unsupervised agglomerative clustering and F-measure values to evaluate sample clustering into biological groups.
Main Results:
- Significant differences in probe set algorithm performance were observed, with 50% showing statistical significance by ANOVA.
- MAS 5.0 p-value weighting significantly improved performance in mouse data (ANOVA and paired t-test).
- The dChip difference model, ProbeProfiler, and RMA algorithms benefited most from p-value weighting. The dChip difference model with continuous p-value weighting demonstrated the best overall performance.
Conclusions:
- Probe set algorithm performance is dependent on the confounding noise levels inherent to specific mRNA profiling projects.
- Optimizing probe set algorithm selection based on project noise levels significantly enhances data analysis.
- Incorporating detection p-value weighting into existing and new probe set algorithms is recommended to improve performance.