Integrative Multidimensional Methods applied to Molecular Classification in Muscle Invasive Bladder Cancer

Abstract

To move toward a molecular taxonomy, we investigated three genome-wide platform data, integrated as an unified categorical code called gene configuration. We performed a multidimensional analysis of the gene configuration on known protein coding genes for 218 patients of muscle invasive bladder cancer, revealing a unified classification into three major subtypes. The classification obtained allow us to describe heterogeneous groups among different ethnic origins. This classification follow mechanistic differences for the different pathways involved in the bladder cancer profiling. A feature selection method based in multiple correspondence analysis allowed to mine the determinant gene configurations involved in the molecular profiling of muscle invasive bladder cancer.