Lin et al. built the callset in two tiers. A baseline set of euchromatic SVs was established from 293 near-T2T haplotype-resolved assemblies generated by HGSVC and HPRC, and this was expanded with an additional 445 low-pass Oxford Nanopore genomes (Schloissnig et al. 2025) and 480 roughly 30x Oxford Nanopore genomes sequenced by the 1000 Genomes Long Read Sequencing Consortium (383 new plus 97 from Gustafson et al. 2024), for 1,218 genomes of diverse ancestry. Structural variants were discovered with ten long-read callers, and a machine-learning tool, BoostSV, ranked and selected the best allele to represent each SV across the different platforms and coverages, producing a single nonredundant callset. The final callset contains 614,522 SVs (376,117 insertions and 238,405 deletions) on T2T-CHM13 and 587,779 SVs (391,410 -insertions and 196,369 deletions) on GRCh38. See Lin et al. for full details. +insertions and 196,369 deletions) on GRCh38. A conventional +linear +callset from HPRC year 2, produced with standard SV callers rather than the +pangenome graph, was used only for comparison and did not contribute to this +merged callset. See Lin et al. for full details.
The insertion/deletion callset VCFs (GRCh38 and T2T-CHM13 native), already annotated with overall and per-superpopulation allele frequencies (EUR, AMR, EAS, AFR, SAS), were provided by the laboratories of Evan Eichler and Danny Miller (University of Washington). At UCSC the deletion and insertion records were converted to bigBed; no re-merging or re-annotation was performed. The step-by-step build commands (format conversion and bigBed build) are recorded in the UCSC makeDoc for this track container: doc/hg38/lrSv.txt. The conversion script and autoSql schema live in makeDb/scripts/lrSv, and the track configuration is in trackDb/human/lrSv.ra.