Detection of Differentially Expressed Groups of Genes Using a Multivariate Test Statistic in a Partially Paired Microarray Data Set

• Activity Number: 387
• Type: Invited
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Memorial
• Author(s): Byung Soo Kim*+ and Johan Lim and Jayeon Kim and Sang-Cheol Kim and Donghyeon Yu and Kyunga Kim
• Companies: Yonsei University and Seoul National University and Yonsei University and Yonsei University and Seoul National University and Seoul National University
• Keywords: partially paired microarray data ; false discovery rate ; correlation among genes ; differentially expressed groups of genes

Abstract

A growing number of reports on the inter-gene correlations in microarray data sets indicate the importance of identifying differentially expressed (DE) groups of genes. We develop a multivariate test procedure for detecting DE groups of genes in a partially paired cDNA microarray data set. The multivariate statistic proposed in this talk, which is referred to as Tp statistic where p stands for the group size of a gene set, generalizes the two-sample Hotelling's T2 statistic. The test based on the Tp can be used for validating the co-expression of preselected gene sets, or for an exploratory search of gene sets whose expression pattern change concordantly between different biological conditions. Once the empirical null distribution of the Tp statistic is derived, we can employ Efron's recent result of modifying the FDR (Efron. 2007) to detect DE gene sets using the conditional FDR.