Equivalence testing to find differentially expressed genes in two-color microarrays

Date
2010
Authors
Seaman, Ronald L.
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract

Equivalence testing was used to identify and rank differential expression of genes represented in microarray data from an experiment on HL60 cells and a type of modified HL60 cells. This type of statistical analysis has previously only been used to find equivalent expression in microarray data. Equivalence of gene expression was tested here using two one-sided t-tests (DOSTT, for double one-sided t-tests) and an equivalence interval between two-fold increase and two-fold decrease in expression. Genes having the largest p-values in equivalence testing were considered the least equivalent and then taken as genes with the most significant difference in expression. Genes were then ranked from largest to smallest p-value to form lists of the genes most likely differentially expressed, or top genes. The performance of the DOSTT method was evaluated by comparing the number of top genes in common with top genes found by four other methods and by comparing the ranks of top genes with ranks assigned by the other methods. A degree of correspondence was found between DOSTT results and results from both a moderated t-test method (limma) and the statistical analysis of microarrays (SAM). Top genes found by DOSTT, limma, and SAM methods were similar to each other in many respects, but results from each one of these methods were different than results from fold change and t-test methods. Results are discussed in terms of the reasons for similarities and differences among methods and are compared with published results of other studies. Based on results of this study, equivalence testing seems able to identify top genes in microarray data as well as they are identified by at least two commonly used methods.

Description
This item is available only to currently enrolled UTSA students, faculty or staff.
Keywords
differential expression, equivalence testing, microarray, top genes
Citation
Department
Management Science and Statistics