Multi-sample variant display

A VCF can hold genotypes for many samples. The multi-sample variant display draws one row per sample, and Show... → Show as genotype matrix in the track menu chooses how the variants run across it:

At genomic positions, the default: each variant drawn across the bases it covers
Equal-width columns with it checked, one column per variant in view, like a heatmap

At genomic positions: structural variants at their real span

Each variant is drawn at its real genomic position. This is the layout that renders structural variants at the right scale, and overlapping calls use slight transparency so you can still tell them apart.

If overlaps overwhelm the view, use "Edit filters" in the track menu to hide variants by size, name, or any Jexl expression.

1000 Genomes SV ensemble callset (3202 samples) across 5 Mb of chr19, one row per sample, sorted by genotype at a 1.1 Mb inversion. Each call is drawn at its real span, so the sort collects the inversion's samples with the inversion into a block against the rest of the cohort.

Equal-width columns: SNP and indel patterns

Each visible variant gets one column and each sample gets one row, regardless of how far apart the variants are on the genome. A thin black line connects each column to its real genomic position.

Sparse small variants that would be only 1–2px wide at their true positions each get a full readable column. Patterns like shared haplotypes, runs of homozygosity, and population structure become visible at a glance.

A phased trio in equal-width columns: one column per variant, one row per haplotype, each cell shaded reference against alt. Inherited haplotype blocks read as contiguous vertical bands shared across parent and child rows.

Filtering by allele frequency and missingness

Two inline sliders in the track menu thin a dense callset down to the variants worth looking at. Both live under Track menu → Filter by... and re-fetch as you release the slider:

Minor allele frequency hides variants whose minor allele frequency falls below the threshold, so singletons and near-monomorphic sites drop out and the common, population-informative variants remain.
Missingness hides variants whose fraction of uncalled alleles rises above the threshold. Its default keeps every variant; lowering it drops the poorly-genotyped columns that are mostly missing data.

Missingness is especially useful in equal-width columns, where each variant takes a full column no matter how many of its genotypes are no-calls.

One tetraploid potato multi-sample VCF opened twice in the same view as a genotype matrix, on one ruler. Top: the default missingness ceiling keeps every variant, and no-call columns dominate. Bottom: a 0.1 ceiling leaves the well-genotyped sites.

Either filter can be preset so the track loads already filtered, with the minorAlleleFrequencyFilter and maxMissingnessFilter display slots. See configuring default display settings.

Rows: per sample or per haplotype

Both layouts color each genotype cell, and how they color it depends on what one row stands for: the unit display option, which the track menu's Rows item sets.

With Rows → Per sample (unit: 'sample'), one cell is drawn per sample and shaded by how many alternate alleles the call has:

homozygous reference (0/0) → light grey
heterozygous (0/1) → a medium shade
homozygous alternate (1/1) → the darkest shade

so the cell color reads directly as allele dosage (0, 1, or 2 alt alleles) and runs of homozygous-alt samples stand out as the darkest blocks. The dosage counts every non-reference allele, so 1/2 draws as dark as 1/1, and an uncalled genotype (./.) takes the no-call color.

With Rows → Per haplotype (unit: 'haplotype'), each sample is split into one row per haplotype and every haplotype cell is colored reference vs alt on its own, with no dosage shading, so inherited haplotype blocks line up as the contiguous vertical bands in the trio matrix above. A haplotype with the site's most frequent alt takes the same blue as a homozygous-alt cell in per-sample rows, and any other alt draws red.

Per-haplotype rows draw only what the VCF phases: |-separated genotypes, and haploid calls, which have nothing to phase. A genotype written with / fills every haplotype row of its sample black, homozygous ones included, since the file does not assign its alleles to haplotypes. Read-backed phasers such as WhatsHap leave homozygous calls unphased, so on their output the black marks every call the phaser left alone, homozygous ones among them. The menu item stays disabled until the view holds phased or haploid genotypes.

Coloring by consequence impact (SnpEff/VEP annotations)

If the VCF's INFO field has SnpEff ANN or VEP CSQ annotations, each variant's alt-allele cells can be colored by the severity of its most severe predicted consequence. From the track menu, open Color by... and choose Consequence impact under Cells. The entry stays visible where no variant in view is annotated, but is greyed out and names why: (checking for SnpEff/VEP annotations...) while the variants load, then (no SnpEff/VEP annotations in view).

Every annotation is bucketed into one of four impact tiers and painted with a fixed color, so the legend is the same across tracks:

HIGH (red) - e.g. stop_gained, frameshift_variant, exon_loss_variant
MODERATE (orange) - e.g. missense_variant, inframe_deletion
LOW (yellow) - e.g. synonymous_variant, splice_region_variant
MODIFIER (grey) - e.g. intron_variant, intergenic_region

This works for both SNVs/indels and structural variants. SnpEff's SV-specific consequence terms (exon_loss_variant, transcript_ablation, gene_fusion, ...) map onto the same four tiers, so a deletion that removes an exon reads as HIGH the same way a stop-gained SNV does.

1000 Genomes phase 3 chr1 genotypes colored by consequence impact from SnpEff annotations: red is stop-gained or splice-site, orange missense, yellow synonymous, and grey the intronic and intergenic majority.

To have the track load already colored this way, set the display's color to { "field": "impact" }: see coloring cells by the variant.

Coloring by SV type

Structural variants can be colored by their class: each alt-allele cell takes the color of its variant's structural-variant type. From the track menu, open Color by... and choose SV type under Cells. Like the consequence option it greys out rather than disappearing, reading (no structural variants in view) where the variants in view hold none.

Each class has one fixed color, the same on every display that paints SV type, so a deletion is the same red here, on the single-variant display, on a mark display and on the SV inspector's chords: deletion (red), duplication (blue), insertion (purple), inversion (orange), copy-number variable (dark brown), tandem repeat (pink), breakend (cyan), complex (green), and other / mixed (black) for a token no class names or a record whose alleles disagree. A het cell draws its class color paler, as in the default coloring, and the legend shows both shades. The legend lists only the classes present in the loaded region, and a record with no structural class keeps the default alt color.

The class is read from the ALT allele first: <DEL>, <DUP:TANDEM>, breakend notation, and 1000 Genomes' <CN0> (a deletion) and <CN2> and up (a duplication). Otherwise it comes from INFO/EVENTTYPE or INFO/SVTYPE, or from a sequence allele 50 bp or more longer or shorter than REF. JBrowse stores the result as each VCF record's svType field.

1000 Genomes SV ensemble callset on chr19 colored by SV type, each alt-allele cell taking its variant's class color, pale for a het and full for a hom. The legend names every class present, including the callset's complex (CPX) events, with both shades beside each.

A track loads already colored by SV type with color set to { "field": "svType" }: see coloring cells by the variant.

Coloring by any field

Color by... → Field... colors each alt-allele cell by a field of its record. The dialog lists the file's columns (QUAL, FILTER, ID, ALT), every INFO field its header declares, with the header's description, and the computed values such as maf and missingness; any other field name or a jexl: expression can be typed in.

Left as it is, each value of the field takes its own color, as INFO.CLNSIG gives Pathogenic, Benign and Uncertain_significance one each.
Cut points turn a number into ranges: 0.001, 0.01 over INFO.AF makes ultra-rare, rare and common bands.

The legend names the field and lists each value or range, and a variant with no value for the field keeps the default alt color. A het cell draws paler than a hom, as in every other coloring.

Coloring and grouping by sample metadata

Samples can be grouped and colored by metadata: population, phenotype, sex, or any attribute you supply. Two slots wire it up:

samplesTsvLocation on the adapter takes a samples TSV whose first column is the sample name (matching the VCF header) and whose every remaining column is a metadata attribute.
rowColor on the display names one of those columns, and the per-sample rows are tinted by that attribute for as long as it is set, over any color the TSV gives a row. Set it back to None to restore the TSV's colors; the arrangement editor's Each row picks a color per sample instead.
facet bands the rows so each value of a column is contiguous: the column's name, or { "field": …, "domain": [...] } with the band order. Each band's name runs up a strip beside the dendrogram where the band is tall enough to hold it. The band is applied over whatever order you have arranged, so dragging a sample into another band snaps it back, and clustering under bands clusters each band on its own, with a dendrogram per band.
rows.domain names the samples that lead the rows; the rest keep the file's order, and a grouping attribute bands within that order.

The JBrowse demo wires up the 1000 Genomes phase 3 chr1 callset (2,504 samples across 26 population codes) this way. For the TSV layout and the adapter and display slots, see auto-coloring samples by metadata.

The 1000 Genomes phase 3 chr1 callset as a multi-sample variant display. All 2,504 samples are sorted and colored by their population code, and each genotype cell is shaded by allele dosage.

You can also change the grouping attribute and colors interactively after the track is open.

Clustering samples by genotype

Samples can be reordered by genotype similarity, via Clustering → Cluster rows by genotype... in the track menu. See Clustering rows for the modes, the dendrogram, and how to share a result in a session URL.

Clustering a multi-sample variant track. Top: the 'Cluster rows by genotype' dialog. Bottom: the rows reordered by genotype similarity, with a dendrogram on the left.

With per-haplotype rows, clustering treats each haplotype as a separate row, so the dendrogram shows haplotype-level relationships.

Multi-sample variant display

At genomic positions: structural variants at their real span#

Equal-width columns: SNP and indel patterns#

Filtering by allele frequency and missingness#

Rows: per sample or per haplotype#

Coloring by consequence impact (SnpEff/VEP annotations)#

Coloring by SV type#

Coloring by any field#

Coloring and grouping by sample metadata#

Clustering samples by genotype#

See also#