A feature ensemble technology to identify molecular mechanisms for distinction between multiple subtypes of lymphoma

Because of complexities and genetic heterogeneities of biological phenotypes, robust computational approaches are desirable to achieve high generalization performance with multiple classifiers, perturbations of the data structures, and biological interpretations. The purpose of this study is to extend our developed ensemble decision approach to distinguish multiple heterogeneous phenotypes and to elucidate the underlying molecular bridges that intertwine the subtypes. Our work identifies the significant molecular mechanisms (disease-relevant genes and functions) that underpin the complex molecular mechanisms for distinction between multiple phenotypes. Feature genes and hierarchical gene cores identified by our method have achieved high accuracy in the classification of multiple phenotypes. The results show that the proposed analysis strategy is feasible and powerful in the classification of biological subtypes and in explaining the molecular connections between clinical phenotypes. Biological interpretations with Gene Ontology revealed concerted genetic pathways for some lymphoma subtypes.