Title: "Mastering the Microarray Mine Field: A Success Story"
Speakers: Drs. J. Romero-Severson, Joe Ogas, Bill Muir; University of Notre Dame, and Purdue University, respectively.
Place: Stanley Coulter (SC) 239; Tuesday, 4:30pm

Abstract

As is often the case with new technology, a heated debate continues on the best way to analyze array data. While all gene expression arrays generate data in which the number of elements (p) greatly exceeds the number of samples (n), all experimental questions are not the same. The problem of p >> n can be simply and effectively handled if the investigator has a one sided hypothesis, pools the error terms and uses orthogonal contrasts for main effects. We also introduce the false discovery number (FDN) to choose the threshold values for the pooled t-tests. The FDN provides investigators with an easily understood risk factor for a given experiment. In the illustrated investigations, many of the genes of interest had near zero values when repressed and low levels of expression when up-regulated. All of the genes selected were tested with qRT-PCR and most of these (87-90%) were confirmed. Our method successfully identified differential expression in many genes labeled "absent" by the chipmaker. The method described will be useful for the identification of genes that determine cellular identity during normal development as well as pathogenesis.


See http://www.stat.purdue.edu/~doerge/BIOINFORM.D/FALL04/sem.html for a full scheule of BIOINFORMATICS SEMINARS.