Download seminar poster
Talk Title :
Which method to use to analyze microarray data?
Date / Time / Location:
Thursday January 15th, 2009 – 6:00 pm
Room S1-151 at IRIC
Since the first days of microarrays, the problem of identifying genes that are significantly over- or under-expressed has kept many biologists, bioinformaticians and statisticians busy. These efforts have brought us an ever growing number of complex statistical approaches but no clear indication on which is the best to use. I will present a novel strategy that we applied to assess the accuracy of 835 distinct microarray analysis pipelines. This strategy exploits the presence of a biological co-regulation network which induces strong dependencies between expression levels observed in microarray data. The accuracy of a pipeline can then be assessed by measuring how well it produces a biologically relevant outcome: one that follows the expected pattern of dependencies. Our results suggest a number of counterintuitive changes in the way microarray data are currently analyzed and challenge well established concepts.