SV caller accuracy assessment. (a) Assessment of SV caller performance on the HG002 GIAB draft T2T assembly-based benchmark for HiFi WGS data from HG002 at 32-fold coverage. Results are stratified by SV size, showing consistently improved F1-score for sawfish across a range of SV sizes. (b) HG002 SV caller performance assessed as in part (a) but for all SV sizes with depth levels subsampled down to 10-fold coverage, showing improved F1-score for sawfish at all coverage levels. (c) Assessment of SV caller joint-genotyping on samples from CEPH Pedigree 1463 generations 2 and 3. For each SV caller, the number of SVs where genotypes are concordant with the pedigree inheritance pattern across all samples are shown compared to the percentage of concordant SVs. “sawfish HQ” shows sawfish results filtered for genotype quality (GQ) ≥40 in all samples. The results show that sawfish calls thousands more concordant SVs than other SV callers while at the same time calling proportionally more concordant SVs, where the proportion of concordant SVs can be increased to over 87% with modest quality filtration that maintains a very high concordant SV count.