Predicting survial by cancer pathway gene expression profiles in the TCGA

Personalized medicine is usually based on known subcategories of a disease for better treatment. Identifying biomarkers that predict disease subtypes has been an important topic in biomedical sciences. There is a controversy as to the optimal number of genes as an input of a feature selection algorithm. In this paper, we investigate the feasibility to use genes pre-selected by biological knowledge rather than all available genes as an input for a feature selection algorithm predicting survival in the glioblastoma of the The Cancer Genome Atlas (TCGA). We discuss the advantage and disadvantage of this approach.