Gene Selection in Classification Problems via Projections onto a Latent Space

The analysis of gene expression data involves the observation of a very large number of variables (genes) on a few units (tissues). In such a context conventional classification methods are difficult to employ both from analytical and interpretative points of view. In this work a gene selection procedure for classification problems is addressed. The dimensionality reduction is based on the projections of genes along suitable non gaussian directions obtained by Independent Factor Analysis (IFA). The performances of the proposed gene selection procedure are evaluated in the context of both supervised and unsupervised classification problems and applied to different real data sets.