Fig. 2From: Optimizing Nanopore sequencing-based detection of structural variants enables individualized circulating tumor DNA-based disease monitoring in cancer patientsDetection of somatic SVs with the SHARC pipeline based on high and low coverage nanopore data. High coverage nanopore sequencing data from COLO829 (melanoma cell line) and HGS-3 (ovarian cancer organoid) were subsampled to low coverages. Outer circles represent the high coverage sets (59x for COLO829 and 56x for HGS-3) and inner circles represent low coverage subsets (4x, 3x, 2x). The following filtering steps were applied in a cumulative manner in the order displayed. a Median percentage of non-somatic (red) and somatic (blue) breakpoints in the raw NanoSV calls for COLO829 (top) and HGS-3 (bottom). b Median percentage of non-somatic (left) and somatic (right) SV calls kept (green) or removed (brown) in the pre-filtering step for COLO829 and HGS-3. c Median percentage of non-somatic (left) and somatic (right) SV calls kept (green) or removed (brown) by the random forest SV classifier for COLO829 and HGS-3. d Median percentage of non-somatic (left) and somatic (right) SV calls kept (green) or removed (brown) by the database filtering for COLO829 and HGS-3. e Median percentage of non-somatic (red) and somatic SV (blue) calls in the complete SHARC output (left) and top 20 largest SVs (right) for COLO829 and HGS-3. f Total number of non-somatic (red) and somatic (blue) SV calls at each step of the pipeline for both COLO829 and HGS-3. In low coverage subsets, all data points are shown and the square box represents the median value. RF, random forest; DBFilter, database filterBack to article page