Developing a label propagation approach for cancer subtype identification problem

Cancer is a disease in which abnormal cells grow uncontrollably and invade other tissues. Several types of cancer have various subtypes with different clinical biological implications. Based on these differences, treatment methods need to be customized. The identification distinct an important problem bioinformatics, since it can guide future precision medicine applications. In order design targeted treatments, bioinformatics attempt discover common molecular pathology subtypes. Along this line, several computational been proposed or stratify into informative However, existing works do not consider the sparseness data (genes having low degrees) result ill-conditioned solution. To address shortcoming, paper, we propose alternative unsupervised method patients using applied numerical algebra techniques. More specifically, label propagationbased approach somatic mutation profiles colon, head neck, uterine, bladder, breast tumors. We evaluated performance our by comparing baseline methods. Extensive experiments demonstrate that highly renders tumor classification tasks largely outperforming state-of-the-art supervised approaches.