Fig. 2:

Somatic benchmarks. a COLO829T/BL tumour and blood cell lines were sequenced in triplicate to 100x/40x. In-silico purity downsampling was performed at 40x normal and 60x tumour coverage. Results are compared against a PCR validated somatic truth set generated from multiple sequencing technologies. b Simulation of somatic breakpoints from the CHM13 telomere to telomere assembly against hg38. Single breakend variant calling allows sensitive detection of breakpoints to satellite repeat regions such as centromeres without increasing the false discovery rate. c GRIDSS2/Manta validation results on 13 patient samples for 50 bp + events. d GRIDSS2/Manta/strelka validation results on 13 patient samples for small (32–100 bp) duplications. e GRIDSS2/Strelka validation results for 32–50 bp events. f Per sample counts of 32–100 bp somatic tandem duplications in the Hartwig cohort. These mutations are enriched in colorectal cancer and associated with ATM driver mutations. g Size distribution of small (32–100 bp) tandem duplications across the Hartwig cohort. This is a distinct signature not associated with microsatellite expansion