Fig. 1.

Comparison of inferred mutation histories based on single-cell and bulk-sequencing data. a Ground truth clonal tree with mutations M₁, …, M₁₀: the coloured nodes represent the subclones, and the tree structure indicates the partial temporal order in which the subclones emerged (from top to bottom). Each subclone contains the mutations it acquired in comparison with its parent and is annotated with the mean VAF of these mutations. (For a heterozygous mutation in a copy-number-neutral region, the VAF is half the mutation’s cellular prevalence.) b Cell mutation profiles obtained from the SCS data for nine cells c₁, …, c₉: ‘1’ indicates the observed presence of a mutation, and ‘0’ absence. A ‘?’ indicates a missing data point (NA), e.g., due to insufficient coverage. The red ‘0’s are false negatives (e.g., drop-out events), the red ‘1’ indicates a false positive. Due to these errors, the mutation matrix defines no perfect phylogeny. c Inferred single-cell mutation tree annotated with single-cell placements. Not all cells can be placed, such that their observed mutation profile matches with the mutations acquired along the lineage from the root to their attachment point. The branching point of the ground truth tree is inferred correctly, due to the strong signal that the red/orange and blue/green mutations do not occur in the same cell. However, mutation order in linear segments is not reliably inferred from the SCS data; especially in the right branch, a mutation with low prevalence (M₁₀) is placed above a more prevalent mutation (M₉) due to errors in the mutation profiles of cells c₆, c₇ and c₉. d Variant allele frequencies obtained from bulk sequencing. VAF-based clustering of mutations leads to merging of subclones. e, f Both clonal trees are compatible with the VAFs and the clustering inferred in (d). Due to the clustering and incompatible VAFs, the correct branching between the blue and red subclone is not inferred, but in both trees, mutation ordering is consistent with their true prevalences