Analysis and interpretation gene-expression profiles

Gene Selection via Discretized Gene-Expression Profiles and Greedy Feature-Elimination

George Potamias1, Lefteris Koumakis1, and Vassilis Moustakis1’2

With the recent advances in microrray technology [6], the potential for molecular diagnostic and prognostic tools seem to come in reality. The last years, microarray-chips have been devised and manufactured in order to measure the expression-profile of thousands of genes. In this context a number of pioneering studies have been con-ducted that profile the expression-level of genes for various types of cancers such as breast, colon, lymphoma, leukemia and other tumors [8], [11], [15], [18]. The aim is to add molecular characteristics to the classification of cancer so that diagnostic proce-dures are enhanced and prognostic predictions are improved [1]. These studies demon-strate that gene-expression profiling has great potential in identifying and predicting

G.A. Vouros and T. Panayiotopoulos (Eds.): SETN 2004, LNAI 3025, pp. 256–266, 2004.© Springer-Verlag Berlin Heidelberg 2004

Microarray technology targets to identify the genes that are expressed in particular cells of an organism at particular time or, at particular conditions (e.g., disease-states or, disease-types). A microarray is typically a glass (or some other material) slide, on to which DNA molecules are attached at fixed locations (spots). There may be tens of thousands of spots on an array, each containing a huge number of identical DNA molecules (or fragments of identical molecules), of lengths from twenty to hundreds of nucleotides. For gene expression studies, each of these molecules ideally should identify one gene in the genome (even if this is not always possible) [4].