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

Analysis and interpretation of gene-expression profiles, and the identification of respective molecular- or, gene-markers is the key towards the understanding of the genetic basis of major diseases. The problem is challenging because of the huge number of genes (thousands to tenths of thousands!) and the small number of samples (about 50 to 100 cases). In this paper we present a novel gene-selection methodology, based on the discretization of the continuous gene-expression values. With a specially devised gene-ranking metric we measure the strength of each gene with respect to its power to discriminate between sample categories. Then, a greedy feature-elimination algorithm is applied on the rank-ordered genes to form the final set of selected genes. Unseen samples are classified according to a specially devised prediction/matching metric. The methodology was applied on a number of real-world gene-expression studies yielding very good results.