Fig. 4

Best practices workflow to call mosaic variants. a Schematic of the filtering strategies used to call mosaic SNVs using WGS and WES data. b GATK (at different ploidy settings) and Mutect2 were used to call simulated mosaic SNVs at different VAFs (x-axis) and sensitivities (y-axis) in DNA mixing experiments. c Reconstructed cell lineage trees using a cohort of mosaic SNVs (Table S2) present in eleven single-cell datasets from the common brain sample. Indicated are the names of each SNV (SNV1, SNV2, etc.) and the estimated SNV VAFs (from 250× WGS data). d SNVs marking the L1 (x-axis) and L2 (y-axis) lineages show anti-correlated VAFs across multiple brain and tissue samples, suggesting these SNVs differentiate the earliest cell lineages in this sample. Solid line, linear regression of the SNV anti-correlation values across all samples. Shaded areas is the corresponding 95% confidence intervals. Dashed line, linear regression of the SNV anti-correlation values when only brain samples are included in the analysis.