|
Activity Number:
|
418
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #306656 |
|
Title:
|
ANOVA Model-Based Pattern Recognition Technique
|
|
Author(s):
|
Yushu Liu*+ and R. Lakshman Chelvarajan and Thomas Getchell and Subbarao Bondada and Arnold J. Stromberg
|
|
Companies:
|
University of Kentucky and University of Kentucky and University of Kentucky and University of Kentucky and University of Kentucky
|
|
Address:
|
200 Thomas Hunt Morgan Building, Lexington, KY, 40506,
|
|
Keywords:
|
pattern recognition ; two-way ANOVA ; PCA (principal component analysis)
|
|
Abstract:
|
In microarray data analysis, clustering methods such as hierarchical, k-means are commonly used to identify similarities in gene expression profiles that may suggest commonalities in underlying functions. These methods are appropriate when there are no pre-conceived notions about the commonalities. In contrast, pair-wise comparisons are used when commonalities are intuited by investigator. Motivated by this incongruity, we developed a pattern recognition metric based on two-way ANOVA and investigator's experimental insight. This method is more stable and more capable of grouping functionally related genes than is k- means clustering. Code is available in Splus. We also used PCA to resolve differences in the expression profiles for multiple probesets of the same gene. Our methods were applied successfully to identify unique expression patterns for genes expressed by splenic macrophages.
|
