|
Activity Number:
|
379
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #304685 |
|
Title:
|
Normalization of Microarray Data Using Multivariate Methods
|
|
Author(s):
|
Mark Reimers*+
|
|
Companies:
|
Virginia Commonwealth University
|
|
Address:
|
730 E Broad St, Richmond, VA, 23298,
|
|
Keywords:
|
microarray ; multivariate analysis ; negative controls ; normalization
|
|
Abstract:
|
Microarray data often has many parameters (P) and few samples (N). "P >> N" works against most traditional statistical procedures. However two procedures introduced here may take advantage of this situation, by adapting traditional multivariate methods to the "P >> N" situation in order to estimate common factors underlying much of typical variation in microarray data. The first procedure uses multivariate analysis of technical replicates in order to identify common factors; weighted regression then removes the estimated effects of these factors. The second procedure identifies common factors correlated between negative control probes and other probes. Robustness and leverage are concerns addressed for both procedures. These procedures are illustrated on some well-known public data sets and are shown to reduce apparent errors by more than a factor of two over current practices.
|
