Fig. 3.

SPACC applied to the B-cell lymphoma data, example 5 in Results. The shapes of the nodes indicate that the samples are categorized into five classes by the input class labels. The polygons show the subsets obtained by SPACC. The FL, CLL and DLBCL classes are correctly separated. The three Tonsil samples are grouped with either DLBCL or FL. Interestingly, the Blood B/T samples (stars) are divided into three subsets, which correspond to blood T-cells, activated blood B-cells, and resting blood B-cells, respectively. Therefore, SPACC successfully identifies subclasses in the Blood B/T class.