Analysis of gene expression profiles at chromosomal level

Transcriptional profiling of whole genomes using cDNA or oligonucleotide high-density arrays is becoming increasingly popular among the biomedical research community. Although advances in technology and the rapid rise in microarray data availability are leading to new insight into fundamental biological problems, investigators are still confronted with the major problem of upgrading the information content of regulated gene lists obtained from microarray experiments. Indeed, the efficient exploitation of gene expression databases requires not only computational tools for management, analysis, and functional annotation of primary data, but also integrating lists of modulated genes with of other sources of genomic information, such as gene sequence, locus or structural characteristics. In particular, integration between expression profiles and chromosomal localizations could be effective in detecting gene structural abnormalities such as genomic gains and losses and/or translocations. The aim of the present study is to apply computational tools for mapping transcriptional data at chromosomal level and detecting clusters of regionally modulated genes in cancer specimens.Statistical tests and signal processing procedures are used to integrate expression profiles and gene sequence information and identify peculiar regions of modulated expression. In particular, the method is based on the application of a smoothing, coordinate-dependent function (e.g., cubic splines) to a standard transcriptional specificity statistic (e.g., standard F-statistic), commonly used to detect differentially expressed genes.This computational tool has been tested on different microarray data sets obtained from various human tumor samples (e.g., solid tumors and hematological disorders). In particular, the application of chromosomal level analysis to the transcriptional database presented by Bhattacharjee (Bhattacharjee et al., 2001), Armstrong (Armstrong et al., 2002), and Ross (Ross et al., 2003) allowed the detection of regional signals corresponding to known as well as putative loci with high frequent genomic losses and gains or marking translocation events.