On the relevance of technical variation due to building pools in microarray experiments