Figure 2. Structural variants as highly specific markers of tumor DNA in cfDNA.

a) Schematic of workflow illustrated with a translocation between chr8 and chr19 identified in OV-107. b) Distribution of VAFs for SVs and SNVs in baseline samples c) Schematic showing how patient specific error rates are calculated by applying probe sets to off target patients d) average background error rates in duplex, simplex and uncollapsed sequences. Each violin/boxplot is a distribution over SVs/SNVs where each data point is the error rate for an individual patient. Triangles show limit of detection (LOD) defined as 2X the largest observed patient error rate e) Fraction of SNV/SVs that have 0 background ie no read support in incorrect patient f) Mean SV VAF vs Tumor fraction computed from TP53 VAF