Chromosomal patterns of gene expression from microarray data: methodology, validation and clinical relevance in gliomas

We have devised a novel mathematical technique (CHROMOWAVE) based on the Haar wavelet transform and applied it to a dataset obtained with the Affymetrix？ HG-U133_Plus_2 array in 27 gliomas. CHROMOWAVE generated multi-chromosomal pattern featuring low expression in chromosomes 1p, 4, 9q, 13, 18, and 19q. This pattern was not only statistically robust but also clinically relevant as it was predictive of favourable outcome. This finding was replicated on a data-set independently acquired by another laboratory. FISH analysis indicated that monosomy 1p and 19q was a frequent feature of tumours displaying the CHROMOWAVE pattern but that allelic loss on chromosomes 4, 9q, 13 and 18 was much less common.The ability to detect expression changes of spatially related genes and to map their position on chromosomes makes CHROMOWAVE a valuable screening method for the identification and display of regional gene expression changes of clinical relevance. In this study, FISH data showed that monosomy was frequently associated with diffuse low gene expression on chromosome 1p and 19q but not on chromosomes 4, 9q, 13 and 18. Comparative genomic hybridisation, allelic polymorphism analysis and methylation studies are in progress in order to identify the various mechanisms involved in this multi-chromosomal expression pattern.Genes are not randomly distributed and their coordinated expression is regulated by their position along chromosomes [1-3]. New mathematical approaches are therefore needed for the analysis of expression microarrays in order to identify variations in expression of spatially related genes and map them along chromosomes. The relationship between changes in DNA copy number and variations of mRNA expression has been previously investigated (for example [4-7]) but only a few studies have used microarrays to examine large chromosomal abnormalities [7-12].Here we propose a novel mathematical model based on single value decomposition (SVD) and Haar wavelets, named CHROMOWA