Mutation landscape as observed by using SCS. (A) Somatic mutations are shown across 97 unique clones (columns) from 38 unique patient samples. Columns are coded at the bottom based on clinical time point and response status. (B) Mutation order was determined by temporally directed edges when 2 putative driver mutations were identified per sample. Mutation ordering was counted across all samples. The edges infer temporal sequences of mutations, and significance is illustrated by Q values (as discussed in "Methods"). Mutations that are likely to occur early are emphasized with a thick border. The resulting graph illustrates the most likely temporal acquisition of mutations across all samples. (C) Cooccurrence of mutations identified from SCS were assessed by using Spearman correlation matrices across all 97 unique clones. (D) Cooccurrence of mutations is indicated from aggregated per-patient VAFs calculated from SCS data. (E) Cooccurrence of mutations from bulk NGS VAFs obtained from the TCGA is indicated for the Tapestri panel mutations. The plot shows cooccurrence (blue) or exclusivity (red) with color coding and the false discovery rate–corrected statistical significance. (F) Comparison of the number of mutation variants per cell identified by using SCS at diagnosis based on the patient’s eventual response. *P < .1; **P < .05; ***P < .01.