Pangenome (mouse)

A pangenome graph of nineteen mouse genomes shows where the strains differ from one another and by how much. This one is the GRCm39 reference plus eighteen inbred and wild-derived strains from the Mouse Genomes Project, aligned together with minigraph. Two things make it worth reading. The reference is itself one of the strains, which inverts the sign of the best-known variant in it. And nobody has published a locus list for this panel, so the loci have to come out of the graph.

We serve the graph as the same rGFA projections the HPRC pangenome page opens, so the tracks, the adapters and the coarse tier are that page's.

Experimental

The graph view is a beta plugin, and this tutorial covers experimental ideas. Where a step below says the view works a particular way today, the step describes a current limit of the view. We welcome your feedback.

Prerequisites

the GraphGenomeView plugin, loaded the way the HPRC page loads it
htslib (tabix), to query the hosted indexes from the command line

Where the data comes from

Each chromosome comes from one minigraph call over the reference followed by the strains. The reference goes first, which gives it rank 0.

minigraph -cxggs -t "$THREADS" $(tr '\n' ' ' < "chrom/$c/order.txt")

The projections are hosted beside the graph:

the segment and link index: https://jbrowse.org/demos/mouse_pangenome/mouse-mm39-minigraph.segs.bed.gz and https://jbrowse.org/demos/mouse_pangenome/mouse-mm39-minigraph.links.bed.gz
the bubble index: https://jbrowse.org/demos/mouse_pangenome/mouse-mm39-minigraph.bubbles.bed.gz
the allele inventory: https://jbrowse.org/demos/mouse_pangenome/mouse-mm39-minigraph.alleles.bed.gz
the coarse tier: https://jbrowse.org/demos/mouse_pangenome/mouse-mm39-minigraph.tier10000

Preparing your own graph describes what each of those files holds and how a graph produces them.

minigraph writes no path lines, so this graph records no carriage: nothing in it records which strain carries a given allele, and firstSeenIn in the allele file is construction order. Pangenome (cattle) is the panel where path lines recover it.

Nnt: a deletion that appears as an insertion

Start at Nnt. Type chr13:119,440,000-119,600,000. C57BL/6J carries a well-known multi-exon deletion there that abolishes the protein and makes B6J mice glucose intolerant. GRCm39 is C57BL/6J, so the backbone of this graph is the strain with the deletion. The graph therefore shows the deletion as sequence that the other strains carry and the reference lacks, the opposite sign from every description of it.

The Nnt locus on GRCm39: RefSeq genes, the bubbles lane, the allele inventory drawn at each allele's real size, and the rGFA segments, over the same window as an anchored graph. The one large allele sits inside Nnt, and it is an insertion because the reference is the strain that lacks the sequence. Open this view in JBrowse ↗ (large dataset — this session takes a while to load)

An insertion consumes almost no reference, so a plain feature track draws a large insertion and a small one at the same minimum width. The allele lane reads the file as alignments and draws each allele at its real size from its CIGAR.

Finding the loci

Nnt is a locus someone had already written about. The rest of this panel has no published literature to read, so the loci have to come from the graph. The coarse tier records how many segments each bubble holds. Ranking the tier by that count reports where the graph varies most, and intersecting the result with the reference annotation names the loci.

The ranking recovers the beta-defensin cluster, the vomeronasal receptor and Speer families and the immunoglobulin heavy chain locus without a curated list. The densest window it returns is a single bubble inside one intron of Dock2, at chr11:34,516,044-34,560,497:

The densest bubble in the mouse graph that still fits in one cut, found by ranking the coarse tier and named off the reference annotation. The gene lane holds nothing but intron, the bubbles lane is a single row, the allele inventory draws each alternative path at its real size, and the graph carries one label naming the whole cut as a superbubble, with Dock2 pinned under the backbone. Open this view in JBrowse ↗ (large dataset — this session takes a while to load)

Because the ranking reads the graph, the method repeats on a panel nobody has written about yet. generatePangenomeLoci.ts in the genomes.jbrowse.org repo computes the ranking. That repo publishes the derived catalogues at genomes.jbrowse.org/pangenomes so a locus can be opened without building anything.

One bubble, one label

A bubble index has one row per bubble, and each row reports where the graph varies and how much. No index row describes the inside of a bubble, so draw the bubble as a force-directed graph and open it to see its contents.

Type chr11:34,516,044-34,560,497, the Dock2 intron, cut it from the segments track with Launch → Graph genome view (this region) and pick Force-directed layout from the Layout dropdown. The index lists this window as a single bubble, so the whole cut is that bubble. One label names it in the index's terms: a superbubble, with its segment count and the span of its routes. A bubble that fills the whole drawing gets the label and no halo, because a halo around everything would mark nothing.

The Dock2 intron window with the RefSeq genes, the bubbles lane and the rGFA segments above the force-directed graph. The bubbles lane is one row, and the graph's one label names the whole cut as a superbubble. Open this view in JBrowse ↗ (large dataset — this session takes a while to load)

The view pins Dock2 under the backbone, and no exon stretch appears anywhere in the cut. The gene track on the graph and the linear view above it both show that the cut is intron.

A superbubble is the index's name for a bubble too big to type. The label gives a segment count and a route range and no kind, since the row's numbers describe the whole region at once.

Dock2 is also the one window on this page that wants the force layout. The other panels are chains and read better anchored; this one is loops hanging off a backbone, which is the structure a force-directed layout suits. Check the node and edge counts in the graph pane's header before switching.

Open it, and open what is inside

Click the label. The view cuts out the bubble's segments and lays out only those. It then derives bubbles from the popped graph. A backbone node that no edge jumps over is a boundary, and whatever lies between two boundaries is a bubble. Each derived bubble gets a halo and a label, and one of them opens in turn. Part 4 of the HPRC tutorial shows one such level opened under its linear view.

Each level has a button that returns to the level above, so you climb back out in the order you descended. The view types the labels at each level the way the index types a bubble. The type comes from the reference interval a bubble replaces and from the shortest and longest route through it. The layering gives those values for any anchored graph.

The control

Nnt holds one large allele that the other strains carry and the reference lacks. Type chr13:119,440,000-119,600,000 and cut it the same way. Nnt should halo as a plain insertion with nothing to descend into.

The Nnt window force-directed. The large loop halos as one insertion beside the smaller sites along the backbone, Nnt's exons run along the reference nodes with its name pinned under them, and no halo is a superbubble. Open this view in JBrowse ↗ (large dataset — this session takes a while to load)

Check it against the index

The Dock2 halo is one row of the hosted bubble index:

tabix https://jbrowse.org/demos/mouse_pangenome/mouse-mm39-minigraph.bubbles.bed.gz \
  'mm39#0#chr11:34516044-34560497'

The label printed this same segment count and route span. No file holds the bubbles inside it. The view derives them from the popped graph's layering each time the level opens, and discards them when the level closes.

Build it yourself

build_mouse_pangenome.sh downloads the assemblies, extracts one sequence per chromosome renamed to PanSN, runs minigraph per chromosome, concatenates, and projects the files above. It is a long run, most of a day of alignment.

The script writes a README.txt beside the data recording the source, the modifications, the tool versions and the audits that ran. Copy the audits into your own build. The build stops unless the reference path reproduces the reference chromosome lengths, and stops on a duplicate segment id after renumbering. Without these audits, either failure produces a graph with wrong coordinates, and every downstream check passes on it.

References

Li H, Feng X, Chu C. The design and construction of reference pangenome graphs with minigraph. Genome Biology. 2020;21:265. https://doi.org/10.1186/s13059-020-02168-z
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

Feedback on this tutorial is welcome: contact us.

Pangenome (mouse)

Prerequisites#

Where the data comes from#

Nnt: a deletion that appears as an insertion#

Finding the loci#

One bubble, one label#

Open it, and open what is inside#

The control#

Check it against the index#

Build it yourself#

See also#

References#

Prerequisites

Where the data comes from

Nnt: a deletion that appears as an insertion

Finding the loci

One bubble, one label

Open it, and open what is inside

The control

Check it against the index

Build it yourself

See also

References