Tumor clustering using independent component analysis and adaptive affinity propagation

Abstract

Tumor clustering is a powerful method in tumor subtype discovery for more accurately and reliably clinical diagnosis and prognosis. In order to further improve the performance of tumor clustering, we introduce a new tumor clustering approach based on independent component analysis (ICA) and affinity propagation (AP). Particularly, ICA is initially employed to select a subset of genes so that the effect of irrelevant or noisy genes can be reduced. The AP and its extensions, adaptive affinity propagation (adAP), are then used for tumor clustering on the selected genes. © 2014 Springer International Publishing Switzerland.