STATISTICAL ANALYSIS OF HIGH DENSITY OLIGONUCLEOTIDE ARRAYS: A SAFER APPROACH

High-density oligonucleotide arrays allow researchers to measure mRNA transcript abundance for thou- sands of genes on a single array. The large number of genes, multiple sources of variation, and typically small number of experimental units (EUs) combine to make analysis of data from these arrays challenging. We describe our experience in applying data analytic techniques to replicated microarray experiments. In particular, we focus on a framework that includes at- tention to Scale, Additive Fits to account for probe and chip biases, assessment of Experimental-unit variability, and where possible, using processes that are Resistant to extreme values (hence the SAFER acronym). Our approach aims to provide results in terms that are familiar to the analyst and easily in- terpreted by biologists. Moreover, by emphasizing resistant methods and not ignoring variability due to sampling, the approach aims to provide measures that speak to the repeatability of the results.