The aCGH technique simultaneously evaluates up to thousands of genomic loci (represented by BAC clones or oligonucleotides) by comparing hybridization dosages of a test DNA and a reference DNA labelled with different dyes, providing genotypic data of putative structural copy number changes within entire genomes. In spite ofmuch effort has been devoted to the pre-processing of aCGH data thereis nothing about how consistent are the results and how could improve their quality. The goal of this work is to derive a statistical model that adequately, define the main systematic sources of artefactual variation involving a single BAC or even an incomplete portion of a BAC and propose the best statistical methods to identify and correct them while maintaining the true biological variations even for single. In that sense, we have developed some useful models based on analysis of variance with a high-order level of cross and nested factors in order to detect various sources of systematic variation poorly corrected bynormalization methods. Some of them have been described, being dye bias (DB), the preferential incorporation by a probe of one specific dye over the other, one of the most important.

array CGH, dye bias, experimental design.