Skip to main content

Hi-C track

TL;DR: a HicTrack with a HicAdapter needs only the .hic file location. Loop and interaction calls (BEDPE) are a separate VariantTrack drawn with a LinearPairedArcDisplay.

Example Hi-C track config:

{
  "type": "HicTrack",
  "trackId": "hic",
  "name": "Hi-C Track",
  "assemblyNames": ["hg19"],
  "adapter": {
    "type": "HicAdapter",
    "uri": "https://jbrowse.org/genomes/hg19/intra_nofrag_30.hic"
  }
}
jbrowse add-track https://jbrowse.org/genomes/hg19/intra_nofrag_30.hic \
  --trackId hic \
  --name "Hi-C Track" \
  --assemblyNames hg19

HicAdapter config

The HicAdapter needs only the .hic file location, given here with the uri shorthand. The longhand form uses a hicLocation slot. See the HicAdapter config docs for all options.

Color scheme

The display's slots for the coloring and its controls:

See the LinearHicDisplay config docs for the full list, and adjusting the color scale for what each one does to the picture.

Normalization and resolution

selectedNormalization names the matrix-balancing scheme to request (KR, SCALE, VC, VC_SQRT, NONE). It is resolved at runtime against what the file actually provides, so a config naming a scheme the file lacks falls back rather than erroring — see normalization.

resolutionBias is a signed offset from the zoom-derived binsize, so the choice stays meaningful as the view zooms: negative is finer, positive coarser.

{
  "type": "HicTrack",
  "trackId": "hic_kr",
  "name": "Hi-C (KR, log scale)",
  "assemblyNames": ["hg19"],
  "adapter": {
    "type": "HicAdapter",
    "uri": "https://jbrowse.org/genomes/hg19/intra_nofrag_30.hic"
  },
  "displayDefaults": {
    "selectedNormalization": "KR",
    "useLogScale": true,
    "showLegend": true
  }
}
jbrowse add-track https://jbrowse.org/genomes/hg19/intra_nofrag_30.hic \
  --trackId hic_kr \
  --name "Hi-C (KR, log scale)" \
  --assemblyNames hg19 \
  --displayDefaults '{"selectedNormalization":"KR","useLogScale":true,"showLegend":true}'

Loops and interactions as arcs

BEDPE loop/interaction calls load as a VariantTrack with a LinearPairedArcDisplay. color is jexl-evaluated per feature and lineWidth sets the arc stroke width in pixels (also draggable from the track menu). This example draws only the high-scoring calls, in dark red, as thin arcs:

{
  "type": "VariantTrack",
  "trackId": "hic_loops",
  "name": "Hi-C loops",
  "assemblyNames": ["hg38"],
  "adapter": {
    "type": "BedpeAdapter",
    "uri": "https://example.com/loops.bedpe.gz"
  },
  "displays": [
    {
      "type": "LinearPairedArcDisplay",
      "color": "jexl:feature.score>=500?'#8b1a1a':'rgba(0,0,0,0)'",
      "lineWidth": 1
    }
  ]
}

Compartments and subcompartments

The compartment eigenvector is a BigWig and needs nothing beyond its location. The bicolor rendering around zero is what a quantitative track does with signed data by default; set minScore and maxScore when two of these tracks have to be read against each other, so neither autoscales to its own extremes:

{
  "type": "QuantitativeTrack",
  "trackId": "hic_compartments",
  "name": "Compartment eigenvector",
  "assemblyNames": ["hg38"],
  "adapter": {
    "type": "BigWigAdapter",
    "uri": "https://encode-public.s3.amazonaws.com/2021/10/28/5b488af0-df49-4b9b-9feb-8ad671b7eaef/ENCFF661LPK.bigWig"
  },
  "displayDefaults": {
    "minScore": -0.03,
    "maxScore": 0.03
  }
}
jbrowse add-track https://encode-public.s3.amazonaws.com/2021/10/28/5b488af0-df49-4b9b-9feb-8ad671b7eaef/ENCFF661LPK.bigWig \
  --trackId hic_compartments \
  --name "Compartment eigenvector" \
  --assemblyNames hg38 \
  --displayDefaults '{"minScore":-0.03,"maxScore":0.03}'

Subcompartments are a BED whose color column carries the class color. ENCODE's copies are not tabix-indexed, so they load with the plain BedAdapter, which reads the whole file — a few hundred kilobytes here:

{
  "type": "FeatureTrack",
  "trackId": "hic_subcompartments",
  "name": "Subcompartments",
  "assemblyNames": ["hg38"],
  "adapter": {
    "type": "BedAdapter",
    "uri": "https://encode-public.s3.amazonaws.com/2022/08/26/7165fc3e-f186-4fba-be87-f4ea600404b0/ENCFF247IAA.bed.gz",
    "columnNames": [
      "chrom",
      "chromStart",
      "chromEnd",
      "name",
      "score",
      "strand",
      "thickStart",
      "thickEnd",
      "itemRgb",
      "numAltClusterings",
      "altClusterNum",
      "altClusterAssignment"
    ]
  }
}
jbrowse add-track-json '{
  "type": "FeatureTrack",
  "trackId": "hic_subcompartments",
  "name": "Subcompartments",
  "assemblyNames": ["hg38"],
  "adapter": {
    "type": "BedAdapter",
    "uri": "https://encode-public.s3.amazonaws.com/2022/08/26/7165fc3e-f186-4fba-be87-f4ea600404b0/ENCFF247IAA.bed.gz",
    "columnNames": [
      "chrom",
      "chromStart",
      "chromEnd",
      "name",
      "score",
      "strand",
      "thickStart",
      "thickEnd",
      "itemRgb",
      "numAltClusterings",
      "altClusterNum",
      "altClusterAssignment"
    ]
  }
}'

columnNames is doing two jobs here. The file has twelve columns whose last three are not BED12's block fields, so under the positional BED layout the cluster count is taken for a blockCount and every feature grows a row of nonexistent subfeatures. Naming column nine itemRgb is what makes the classes paint their own colors: the file's own header spells it itemRGB, which is not the name JBrowse looks for, and the header would not be consulted anyway because a second comment line follows the column line — the parser takes the last header line as the definition, and that one has no tab-separated fields. Neither failure raises an error.

See also