Figure 2: Treeness tests on ENCODE data.

(a) Distribution of α-values (significance values) of the treeness test for all tetrads from the 12 normal cell types in the ENCODE data. The high mode near α=0 indicates that there is a large number of tetrads where the null hypothesis is false, that is, where the tetrad has true tree structure. From this distribution, one can calculate that 100*(1−π₀)=67% of the tetrads have true tree structure. (b) Tree relationship among normal cell types based on discretized transcriptomic data using maximum parsimony method²¹. Colours indicate different groups of cells. Magenta: fibroblast and myoblast cells; cyan: immune cells; orange: epithelial cells; black: others. (c) Neighbour net^22,23 for the distance matrix for normal cells. Scale bar, 1,000. (d) Neighbour net of the distance matrix for neoplastic cells. Scale bar, 1,000. Note that the cancer cells have a greatly less resolved tree than normal cells, which is also reflected in the treeness statistic. Colour-marked samples are used for treeness test among cancer origin pairs.