Skip to main content
Fig. 8 | Algorithms for Molecular Biology

Fig. 8

From: A multi-labeled tree dissimilarity measure for comparing “clonal trees” of tumor progression

Fig. 8

Clonal trees of tumor evolution, inferred by SiFit and PhISCS, for triple-negative breast cancer (TNBC) dataset originally published in [37] and consisting of the binary presence/absence profile of 22 mutations across 16 single cells. Names of the clones are assumed not to be included as part of the vertex label. Trees are very similar to each other in placement of the vast majority of mutations: (i) Clone 1 in the SiFit tree is almost identical (with respect to the set of mutations assigned to its label) to Clone 1 in PhISCS tree (ii) Clone 2 in SiFit tree is split into two adjacent clones, namely Clone 2 and Clone 3, in PhISCS tree. Analogous applies to Clone 7. (iii) The order of mutations in genes CHRM5 and TGFB2, as well as in most other pairs of mutations (including the pairs where both mutations are at the same vertex), is same among the trees. Notable exceptions leading to some dissimilarities between the trees include mutations in genes MAP3K4 and ECM1. In addition, mutations in genes CBX4 and TNC are absent in tree reported by SiFit. Removing these four mutations and their corresponding vertices from each tree (if present) and assigning each of the Clone 4 and Clone 7 in SiFit tree as child of Clone 2, and Clone 7 as child of Clone 3 in PhISCS tree, we obtain trees which are same up to the existence of splits of single into two adjacent clones belonging to the same lineage (see (ii) from above). MLTD-normalized score for the two trees equals 0.82, which well reflects the overall high topological similarity and concordance in ordering pairs of mutations

Back to article page