|
Activity Number:
|
233
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #305297 |
|
Title:
|
Using Hierarchical Group Filters in High-Throughput Analysis to Improve the True Positive Discovery Rate
|
|
Author(s):
|
Shesh N. Rai*+ and Christopher N. Barnes
|
|
Companies:
|
University of Louisville and University of Louisville
|
|
Address:
|
Delia B. Baxter Biomedical Research Building, Louisville, KY, 40292,
|
|
Keywords:
|
filtering ; high-throughput ; microarray
|
|
Abstract:
|
The problem of biomarker selection in high-throughput experiments has grown as the size of the experiments has grown. Many methods have been developed to control the false discovery rate (FDR), but few have focused on improving the true discovery rate (TDR). Filtering methods, which diminish the multiplicity effect, are one such method; though many of the filters are biased in construction. We propose two unbiased group filters based on hierarchical levels of fold change and p-values that are robust to a wide range of differential expression and normality assumptions. We demonstrate through simulation that the group filters improve the TDR while maintaining strong control of the FDR compared to FDR methods alone. The approach was favorably utilized for biomarker detection in a melanoma study when use of FDR alone methods did not produce substantial results.
|
