are microarray data reliable?
The use of the Affymetrix whole genome ATH1 array (Redman et al., 2004) results in over 20 000 expression data, which have to be processed to estimate the reliability of each single data point in terms of effective expression of the gene and evaluation of the statistical significance of a change of expression when comparing two datasets (e.g. RNA from an aerobic sample with that of an anaerobic sample). Besides the importance of a careful analysis of the results arising from each experiment, the need for a validation of results by comparing experiments performed by independent researchers is often under-evaluated. Indeed, one expects that the microarray technology not only provides a wealth of data about genes induced or repressed by low O2, but also that the genes identified (or at least a large proportion of genes) respond to low O2 independently of experimental conditions used to limit O2 availability. In the next paragraph, we evaluate the ability of microarrays to identify low-O2-
