Characterizing Cancer Subtypes using the Dual Analysis Approach in Caleydo

The comprehensive analysis and characterization of cancer subtypes is an important problem to which significant resources have been devoted in recent years. In this paper we integrate the dual analysis method, which uses statistics to describe both the dimensions and the rows of a high dimensional dataset, into StratomeX, a Caleydo view tailored to cancer subtype analysis. We introduce significant difference plots for showing the elements of a candidate cancer subtype that differ significantly from other subtypes, thus enabling analysts to characterize cancer subtypes. We also enable analysts to investigate how samples relate to the subtype they are assigned and to the other groups. Our approach gives analysts the ability to create well-defined candidate subtypes based on statistical properties. We demonstrate the utility of our approach in three case studies, where we show that we are able to reproduce findings from a published cancer subtype characterization.