Segmental duplications were detected with SEDEF, which finds pairs of homologous genomic segments by seeding on shared k-mers and extending and refining the alignments, then reports each duplicated segment pair with a set of alignment statistics (see reference below). The calls were produced by the Eichler laboratory as part of the HPRC assembly annotation.
The annotation files were obtained from the HPRC Release 2 data collection on the public s3://human-pangenomics bucket, indexed at the hprc_intermediate_assembly data tables. The per-assembly SEDEF output was converted to a UCSC bigBed file, keeping the region coordinates, the paralog partner, and the main alignment statistics. The sequence name of the paralogous region is translated from the HPRC PanSN naming to the name shown elsewhere in the browser. The steps are described in the -makeDoc, +makeDoc, the build scripts are in the kent source tree, and the track configuration is in trackDb/contrib/hprc2annot.
Every duplication call in the source files is kept, with one exception. For HG00735 haplotype 2 (GCA_018472765.3) the SEDEF calls were computed against an earlier version of the assembly's contigs, so about 40,000 contig-level calls name sequences that do not exist in the assembly served here and could not be placed; the roughly 21,600 chromosome-level calls are unaffected, although about half of them name one of the missing contigs as their paralogous region, so on that assembly alone the partner is shown in the HPRC naming rather than translated. This is a version mismatch in the source data rather than something introduced in the conversion. It is the only assembly of the 462 where either