# Pangenome (HPRC): repeat lengths across haplotypes

:::caution JBrowse v5 beta

The tutorials target the JBrowse v5 beta, and the v4.3.0 release
on the [download page](https://jbrowse.org/jb2/download/) lacks some of what they show. To install the
beta of JBrowse Web, run `npm install -g @jbrowse/cli@next`, then
`jbrowse create jbrowse2 --branch v5.0.0-beta.9`. Desktop beta builds are coming soon.

:::

A tandem repeat is one sequence copied head to tail, and how many copies a
person carries varies. An assembled haplotype spans the whole array, so the
Human Pangenome Reference Consortium's release 2 graph holds each haplotype's
repeat at its full length. We read two repeats off it, one bar per haplotype:

- at _LPA_, count copies of the kringle IV type 2 repeat
- at _ABCA7_, measure an intronic VNTR and set PacBio's read-based TRGT
  genotypes on the same bars
- find the samples where the assemblies and the reads disagree

:::caution Experimental

The graph view is a beta plugin, and walk rows is its newest layout. We welcome
your [feedback](https://jbrowse.org/jb2/contact).

:::

## Prerequisites

- the GraphGenomeView plugin, which the HPRC page's launches and the hosted HPRC
  config the sessions below open already load;
  [hosting your own graph](https://jbrowse.org/jb2/docs/tutorials/pangenome_prepare_graph#the-graphgenomeview-plugin)
  loads it into your own JBrowse

## Where the data comes from

[HPRC release 2](https://doi.org/10.64898/2026.07.21.739710) and PacBio's TRGT
genotypes of 100 of its samples (Dolzhenko et al. 2024):

- our tabix projections of the release's rGFA graph, with the exact build
  recorded beside them: https://jbrowse.org/demos/hprc/README.txt
- the release 2.1 gbz-base database, one walk per haplotype, read by range
  request:
  https://s3-us-west-2.amazonaws.com/human-pangenomics/pangenomes/freeze/release2/minigraph-cactus/v2.1/hprc-v2.1-mc-grch38/hprc-v2.1-mc-grch38.gbz.db
- our companion index naming that database's haplotypes:
  https://jbrowse.org/demos/hprc/hprc-v2.1-mc-grch38.haplotype-index.anchored.db
- TRGT's genotypes over the Genome in a Bottle repeat catalogue, a TRGTdb:
  https://zenodo.org/records/8329210/files/adotto_hprc.tdb.tar
- the catalogue itself:
  https://zenodo.org/records/8329210/files/adotto_repeats.hg38.bed.gz
- the _ABCA7_ record of those genotypes, as a TRGT VCF:
  https://jbrowse.org/demos/hprc/hprc_abca7_trgt.vcf.gz

## The LPA kringle repeat

_LPA_ carries a tandem array of kringle IV type 2 (KIV-2) copies, tied to
lipoprotein(a) levels, a heart-disease risk factor (Schmidt et al. 2016). Open
the [HPRC page](https://staging.genomes.jbrowse.org/pangenomes/hprc) and press
**graph** on the LPA row. JBrowse opens on `chr6:160,525,000-160,655,000`. Pick
**Layout → Force-directed layout** from the graph track's menu.

<Figure caption="The LPA window with the RefSeq genes, UniProt's kringle domains and the HPRC bubbles above the force-directed graph track. The kringle array is the knot of loops in the middle, haloed and labelled as a repeat array, and LPA is pinned under the backbone with its exons along it." src="/img/pangenome/hprc_lpa_kiv2.png" />

Every loop in the knot is a different number of copies. Click the array's purple
label to lay the bubble's segments out alone, with a button back to the window.
This rGFA cut records only segments and links, so which haplotype takes which
loop stays unanswered here.

## One haplotype's copies

A **walk** is one haplotype's route through the graph, and the release publishes
one per haplotype as a gbz-base database. Type the array's own window,
`chr6:160,616,002-160,646,753`, put the rGFA graph track back to a segments lane
with **Display types → Feature display**, and turn the gbz-base track on from
the track selector. **Display types → Graph** in its track menu then cuts that
same track as a graph of the walks. Enter the eight lanes the track drew in the
**Haplotypes** field of **Settings**,
`HG00097.1, HG00099.1, HG00128.1, HG00133.1, HG01109.1, HG01123.1, HG01960.1, HG02055.1`,
and pick **Layout → Force-directed layout**.

A node draws thicker the more walks carry it (Bandage's depth as width), so the
shared backbone is the thick line and one haplotype's own copies are the thin
loops. Untick **Mark bubbles** so the per-route chips stop stacking, then pick
`HG00133` under **Walk**: its route keeps its ink while everything else fades,
and a readout gives its length against the reference walk.

<Figure caption="The eight-haplotype KIV-2 cut under the same window's genes, bubbles and rGFA segments, with HG00133 picked under Walk. The labelled loop is copies HG00133 walks and GRCh38 does not, its links drawn dark, and the readout states the walk's excess over GRCh38." src="/img/pangenome/graph_kiv2_walks.png" />

Pick **Layout → Walk rows** and **Color → Uniform**: each walk becomes a bar,
longest first, blue where GRCh38 carries the same sequence and purple where it
does not, so the copies a haplotype adds read as its purple stretch.

<Figure caption="The eight-haplotype KIV-2 cut in walk rows, one bar per haplotype under GRCh38's, longest first, under LPA with the KIV-2 bubble boxed in the bubbles lane. The purple stretch of each bar is kringle copies GRCh38 does not carry, and each readout gives the walk's length and its excess over GRCh38." src="/img/pangenome/graph_kiv2_walk_rows.png" />

Click the KIV-2 bubble boxed in the bubbles lane. Its details give
`shortestAlleleLength` and `longestAlleleLength`, the shortest and longest
routes the rGFA holds there. Every bar in walk rows falls between them, and
GRCh38's, with no purple, is the shortest of the nine.

## The ABCA7 VNTR

Each walk's length comes from an assembly, and HiFi reads measure the same
repeat independently. An intron of _ABCA7_ holds a VNTR tied to Alzheimer's
disease risk (De Roeck et al. 2018), which PacBio genotyped from HiFi reads in
100 HPRC samples. Open the session below: the genes, the catalogue's VNTR row
and the TRGT genotypes, over the gbz-base graph track in **Walk rows** layout
with **Uniform** color:

```json session config=https://jbrowse.org/demos/hprc/config.json
{
  "defaultSession": {
    "name": "ABCA7 VNTR, walk rows",
    "sessionTracks": [
      {
        "type": "FeatureTrack",
        "trackId": "abca7_vntr",
        "name": "Tandem repeat catalogue (adotto)",
        "assemblyNames": ["hg38"],
        "adapter": {
          "type": "FromConfigAdapter",
          "features": [
            {
              "uniqueId": "abca7_vntr",
              "refName": "chr19",
              "start": 1049407,
              "end": 1050096,
              "name": "ABCA7 VNTR"
            }
          ]
        }
      }
    ],
    "views": [
      {
        "type": "LinearGenomeView",
        "assembly": "hg38",
        "loc": "chr19:1,049,000-1,050,500",
        "tracks": [
          {
            "trackId": "hg38_ncbiRefSeq_ucsc",
            "type": "LinearBasicDisplay",
            "showOnlyGenes": true,
            "displayMode": "compact",
            "height": 40
          },
          {
            "trackId": "abca7_vntr",
            "type": "LinearBasicDisplay",
            "height": 40
          },
          {
            "trackId": "hprc_abca7_trgt",
            "type": "LinearVariantDisplay",
            "height": 40
          },
          {
            "trackId": "hprc_v2_1_gbz_lanes",
            "type": "LinearGraphDisplay",
            "layoutMode": "walkrows",
            "colorScheme": "uniform",
            "height": 600
          }
        ]
      }
    ]
  }
}
```

Each row is one haplotype's walk between the flanking reference nodes, blue
where it matches GRCh38 and purple where it does not; GRCh38's own walk is the
short bar at the top.

## TRGT's calls on the same bars

[TRGT](https://github.com/PacificBiosciences/trgt) genotypes tandem repeats from
HiFi reads. We host its _ABCA7_ call as this track:

```json addtrack
{
  "type": "VariantTrack",
  "trackId": "hprc_abca7_trgt",
  "name": "TRGT repeat genotypes at ABCA7, 100 HPRC samples",
  "assemblyNames": ["hg38"],
  "adapter": {
    "type": "VcfTabixAdapter",
    "uri": "https://jbrowse.org/demos/hprc/hprc_abca7_trgt.vcf.gz"
  }
}
```

TRGT writes the same fields for your own samples, one VCF each, joined by
`trgt merge`:

```bash
# one run per sample: HiFi reads aligned to GRCh38, and the repeat catalogue
trgt genotype --genome GRCh38.fa --reads sample.bam \
  --repeats adotto_repeats.hg38.bed --output-prefix sample
# one multi-sample VCF, which the steps below read unchanged
trgt merge --vcf *.vcf.gz --genome GRCh38.fa --output-type z --output merged.vcf.gz
```

With the TRGT track in the session, a **Repeat** dropdown appears beside
**Walk**. Pick the _ABCA7_ record: the bars divide into motif-length units, each
walk gets a black tick at the allele TRGT called for it (paired by length), and
a readout turns red past 10% apart.

<Figure caption="The ABCA7 VNTR in walk rows, one bar per HPRC haplotype, longest first, with the TRGT record picked in the Repeat dropdown, boxed. The catalogue lane marks the VNTR on GRCh38, whose walk is the short blue bar at the top. Each bar is tiled by the motif, and a black tick marks the allele TRGT called for that walk. A red readout is a walk far from its allele." src="/img/pangenome/hprc_abca7_repeat_units.png" />

## The samples where they disagree

The session below shows seven samples' walks in pairs, longest first, named in
`walkRowSamples`:

```json session config=https://jbrowse.org/demos/hprc/config.json
{
  "defaultSession": {
    "name": "ABCA7 VNTR, where reads and assemblies disagree",
    "sessionTracks": [
      {
        "type": "FeatureTrack",
        "trackId": "abca7_vntr",
        "name": "Tandem repeat catalogue (adotto)",
        "assemblyNames": ["hg38"],
        "adapter": {
          "type": "FromConfigAdapter",
          "features": [
            {
              "uniqueId": "abca7_vntr",
              "refName": "chr19",
              "start": 1049407,
              "end": 1050096,
              "name": "ABCA7 VNTR"
            }
          ]
        }
      }
    ],
    "views": [
      {
        "type": "LinearGenomeView",
        "assembly": "hg38",
        "loc": "chr19:1,049,000-1,050,500",
        "tracks": [
          {
            "trackId": "abca7_vntr",
            "type": "LinearBasicDisplay",
            "height": 40
          },
          {
            "trackId": "hprc_abca7_trgt",
            "type": "LinearVariantDisplay",
            "height": 40
          },
          {
            "trackId": "hprc_v2_1_gbz_lanes",
            "type": "LinearGraphDisplay",
            "layoutMode": "walkrows",
            "colorScheme": "uniform",
            "repeatTrackId": "hprc_abca7_trgt",
            "repeatKey": "chr19:1049406-1050096",
            "walkRowSamples": [
              "HG00099",
              "HG03688",
              "HG00741",
              "HG02647",
              "HG01943",
              "HG02559",
              "HG04199"
            ],
            "height": 360
          }
        ]
      }
    ]
  }
}
```

HG00099, HG03688 and HG00741 tick at the end of each bar: reads and assemblies
agree. HG02647 and HG01943 turn red: TRGT calls each near-homozygous while the
graph carries a haplotype neither call reaches (two, in HG01943). HG02559 and
HG04199 carry a grey tick: no read spans HG02559's second allele, and HG04199's
assembly does not span the repeat, so neither walk gets a verdict.

<Figure caption="Seven samples' walks through the ABCA7 VNTR in pairs, each bar carrying the allele TRGT called for it as a tick. A tick at the end of its bar is agreement, a red readout is a walk far from its allele, and a grey tick is an allele no read spanned." src="/img/pangenome/hprc_abca7_disagreements.png" />

## Check it against TRGT's genotypes

Click the TRGT record: its sample table gives `AL`, the allele lengths behind
each tick, and `SD`, the reads spanning each allele. HG02559's second allele has
none, which is why its tick is grey.

## Reproduce it end to end

```bash
curl -fO https://raw.githubusercontent.com/GMOD/jbrowse-components/main/scripts/build_hprc_abca7_trgt.sh
bash build_hprc_abca7_trgt.sh       # writes ./hprc_abca7_trgt_build/
```

The TRGTdb is a directory of Parquet tables, and DuckDB reads a locus's calls
straight out of it:

<!-- from: scripts/build_hprc_abca7_trgt.sh -->

```bash
# every sample's allele numbers and spanning-read counts at one locus
duckdb -json -c "
  select regexp_extract(filename, 'sample\.(.*)\.pq', 1) as sample,
    allele_number, spanning_reads
  from read_parquet('hprc_100.tdb/sample.*.pq', filename = true)
  where LocusID = (select LocusID from read_parquet('hprc_100.tdb/locus.pq')
    where chrom = 'chr19' and start = 1049407)"
```

[`build_hprc_abca7_trgt.sh`](https://github.com/GMOD/jbrowse-components/blob/main/scripts/build_hprc_abca7_trgt.sh)
downloads PacBio's TRGT database and the repeat catalogue, and writes the
_ABCA7_ record with every sample's genotype.

## See also

- [](https://jbrowse.org/jb2/docs/tutorials/pangenome_hprc)
- [](https://jbrowse.org/jb2/docs/tutorials/pangenome_hprc_haplotypes)

## References

- Schmidt K, Noureen A, Kronenberg F, Utermann G. Structure, function, and
  genetics of lipoprotein(a). J Lipid Res. 2016;57(8):1339-1359.
  https://doi.org/10.1194/jlr.R067314
- Wick RR, Schultz MB, Zobel J, Holt KE. Bandage: interactive visualization of
  de novo genome assemblies. Bioinformatics. 2015;31(20):3350-3352.
  https://doi.org/10.1093/bioinformatics/btv383
- De Roeck A, Duchateau L, Van Dongen J, et al. An intronic VNTR affects
  splicing of ABCA7 and increases risk of Alzheimer's disease. Acta Neuropathol.
  2018;135(6):827-837. https://doi.org/10.1007/s00401-018-1841-z
- Dolzhenko E, English A, Dashnow H, et al. Characterization and visualization
  of tandem repeats at genome scale. Nat Biotechnol. 2024;42(10):1606-1614.
  https://doi.org/10.1038/s41587-023-02057-3
- TRGT repeat catalogues and HPRC genotypes, Zenodo.
  https://doi.org/10.5281/zenodo.8329210
- [HPRC release 2](https://doi.org/10.64898/2026.07.21.739710)