Cancer is the second leading cause of death worldwide. Yet, we still have little knowledge on it – one reason is that there is no single cancer disease, but the word “cancer” is rather used to refer to any of the 200 diseases that are characterized by an uncontrolled growth of cells, invading and damaging the body’s normal tissues. As no cancer is like the other, researchers strike to identify the main actors of the uncontrolled cell growth. Those main actors are typically genes that are abnormally (higher or lower) expressed in cancer cells and thus negatively affect the cell processes. Researchers are especially interested in changed behavior of gene expressions, as this tells them a lot about the molecular processes and relationships in cancerous cells, e.g gene functions and interactions.

RNA-Sequencing (RNA-Seq) delivers a complete snapshot of gene expression in a cell, with a single experiment containing expression levels of tens of thousands of genes from multiple hundred samples. The nature of gene expression data, however, poses challenges to its analysis in terms of its high dimensionality, noise, and complexity of the underlying biological processes to detect. Researchers aim to identify genes that reliably discriminate sample groups from each other, e.g. a cancerous from a healthy one. The current state-of-the-art for gene selection is to apply traditional statistical and machine learning approaches, e.g. Support Vector Machines (SVM).