Substitutions, indels and structural changes were called with the CaVEMan, Pindel and BRASS algorithms, which are described in detail elsewhere42. In addition to previously reported filtering41, we asked that all variants were unique to each HSC clone and not found in unrelated HSC or tail samples, or in unrelated mouse strains. We did not apply a VAF filter for clonality because when we examined the VAF distribution of the final datasets, these were centered around 0.5 (Extended Data Fig. 9). For the transcriptional analysis, previously reported HSC RNA-seq data was aligned with Bowtie2 (http://bowtie-bio.sourceforge.net/bowtie2/index.shtml against NCBIM3843, and the overlap between indels and transcribed genes was calculated in R. To calculate the fraction of the genome covered by genes, positions of genes were retrieved from Ensembl. Overlapping regions between two genes were taken into account and counted only once.
