Fig. 3

Summary of validation results for 400 candidate mosaic SNVs. Vertical lines represent candidate mosaic SNVs. Shaded rectangles to the right of the figure provide the keys to interpret the shading presented for each candidate SNV. There was concordance in true-positive mosaic SNV calls (PASS; green rectangle at bottom of figure) in multiple datasets and secondary validation experiments. Chromium linked read haplotype phasing and single-cell sequencing datasets also were effective in supporting a subset of bona fide mosaic SNV calls. By comparison, the VAFs of false-positive calls (red rectangle) are inconsistent across different datasets and often occur within or near insertion/deletion (indel) mutations, short tandem repeat sequences (STRs), homopolymeric nucleotide stretches, or copy number variants (CNVs). Importantly, the panel of normal (PON) filter, but not the comparison to WGS data from a control sample (i.e., to NA12878), was highly effective at identifying contaminating false-positive SNV calls (orange rectangle) and germline SNPs (gray rectangle). We lacked sufficient data to evaluate a subset of candidate SNVs (purple rectangle, NED—not enough data). The two green triangles at the top of the figure denote mosaic SNVs that validation experiments deemed to be false-positive calls; however, cell lineage analyses demonstrated that they are likely bona fide mosaic SNVs (see text and Fig. 4)