Fig. 4

Clonal evolution during treatment. Graphical representation of clonal evolution during sequential epirubicin and docetaxel*, for selected patients: a DDP013, b DDP076, c DDP014, and d DDP103. The top panel within each subfigure (a–d) show allelic prevalence of mutation clusters, after clustering using the PyClone algorithm (see Additional file 1), based on variant allele frequencies (VAFs) for all mutations in the samples extracted for each patient. “Clusters” with one mutation have been merged to nearest cluster based on z-score (range of −1, +1) probability. Subsequent clusters with <3 mutations have been removed from the panel, for clarity. The middle panel within each subfigure is a visualization of a likely pattern of tumor evolution using the “timescape algorithm” (see Additional file 1). These models are based on the clusters in the top panels. Diagrams do not distinguish between new subclones appearing, harboring all truncal mutations and those appearing that have lost some truncal mutations. Each color represents an estimated subclone from the mutation clusters. The horizontal axis denotes three different time points during tumor evolution: pretreatment, post-epirubicin, and post-docetaxel. The bottom panel within each subfigure (coxcomb plots) represents somatic aberrations (mutations and copy number alterations; CNAs) for each time point (pretreatment, post-epirubicin, and post-docetaxel). Gray wedges represent merged “passenger mutations” while colored wedges represent driver somatic mutations and driver CNAs. Relative variant allele frequencies (rVAFs) as well as logR are presented by lateral extension of an outlined wedge. *HER2-positive breast cancers received concomitant docetaxel and trastuzumab