Structural variants (Dog10K)
TL;DR: slice a locus out of the Dog10K structural-variant callsets over
HTTP, load it as a VariantTrack in the multi-sample variant display with breed
labels, and read the genotypes against the gene model above it. Four loci, one
recipe, a different class of variant each time.
Prerequisites
- nothing to read along. Everything below is for building the tracks yourself
- the
UU_Cfam_GSD_1.0dog assembly set up in JBrowse (UCSC calls it canFam4, see the assemblies guide) bcftoolsbuilt with libcurlcurlpython3- htslib (
tabix) minimap2, for the FGF4 synteny halfsamtools, for the FGF4 synteny half- the UCSC
liftOverbinary for the OMIA lane, which the build script fetches itself
On Debian/Ubuntu, apt install bcftools samtools minimap2 tabix curl python3
covers the rest. The scripts write local files, which
JBrowse Desktop opens by path and JBrowse Web takes
through Add track.
Where the data comes from
Two Dog10K structural-variant callsets from Schall & Kidd (2025), read directly over HTTP, plus supporting UCSC and OMIA tracks and two sequenced retrocopies from GenBank.
- the Zenodo Paragraph callset, 5.9 GB, carrying the NHEJ1 deletion and the RNASE1 insertion: https://zenodo.org/api/records/14968874/files/Dog10k_manta_paragraph.vcf.gz/content
- the Michigan Manta aggregate callset, 1.08 GB, carrying the AMY2B duplication and the FGF4 intron records: https://kiddlabshare.med.umich.edu/dog10K/Manta-SV_2022-03-28/SV-genotype-v2.merge.agg_only.08032022.vcf.gz
- the sample table, breed and category per animal, behind every panel on this page: https://kiddlabshare.med.umich.edu/dog10K/sample-information/dog10K-alignment-sample-table.2022-02-23-v7.txt
- OMIA's own dump, curating the Collie eye anomaly record independently of either callset: https://omia.org/static/omia.sql.gz
- the canFam3-to-canFam4 chain that lifts OMIA's coordinates: https://hgdownload.soe.ucsc.edu/goldenPath/canFam3/liftOver/canFam3ToCanFam4.over.chain.gz
- the
UU_Cfam_GSD_1.0gene annotation, checking the FGF4 records against the gene's introns and drawing the parent-gene track in the synteny figure: https://jbrowse.org/ucsc/canFam4/ncbiRefSeq.gff.gz - the CFA18 retrocopy, MF040222, fetched from GenBank: https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=nuccore&id=MF040222&rettype=fasta&retmode=text
- the CFA12 retrocopy, MF040221, fetched from GenBank: https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=nuccore&id=MF040221&rettype=fasta&retmode=text
- the FGF4 parent-locus sequence the two retrocopies are aligned against, over UCSC's canFam4 REST API: https://api.genome.ucsc.edu/getData/sequence?genome=canFam4;chrom=chr18;start=48865000;end=48876000
A 7.8 kb deletion in NHEJ1
Schall and Kidd genotyped long-read-discovered structural variants across the Dog10K collection and flagged those whose allele frequencies track breed clades. One is a 7.8 kb deletion in an intron of NHEJ1, the variant Parker et al. (2007) tied to Collie eye anomaly. It should be common in Collies and their relatives and absent from unrelated breeds and wolves. The anomaly is recessive, so the darker cells below are affected animals and the lighter ones carriers.
Slicing one locus out of the callset
The genotype VCF is 5.9 GB across 1,879 dogs and wolves, published on
Zenodo with a tabix index, and
bcftools fetches only the locus. Zenodo serves the data and index from
separate URLs, so the index is named explicitly:
Z=https://zenodo.org/api/records/14968874/files
SV=$Z/Dog10k_manta_paragraph.vcf.gz/content
SVI=$Z/Dog10k_manta_paragraph.vcf.gz.tbi/content
bcftools view -r chr37:25500000-25620000 -S sv.samples --force-samples \
-Oz -o dog10k_nhej1_svs.vcf.gz "$SV##idx##$SVI"
tabix -p vcf dog10k_nhej1_svs.vcf.gzsv.samples comes from the Dog10K sample table: every Collie, Shetland Sheepdog
and Silken Windhound in the analysis set, four Lancashire Heelers, Australian
Shepherds, German Shepherds and Labrador Retrievers as breeds with no reported
association, and four Greek gray wolves as the outgroup. Read the genotypes
directly first:
bcftools query -r chr37:25574005-25574006 -f '[%SAMPLE=%GT ]\n' \
dog10k_nhej1_svs.vcf.gz | tr ' ' '\n' | grep -v '=0/0'Eleven of the thirteen Collies carry it, four homozygous, along with two of four Shetland Sheepdogs and one of two Silken Windhounds. Nothing else carries a copy.
Loading the slice with breed labels
An SV VCF loads as an ordinary VariantTrack. The multi-sample variant display
draws one row per sample across the variant's genomic span, so a 7.8 kb deletion
is a 7.8 kb block.
{
"type": "VariantTrack",
"trackId": "dog10k_nhej1_svs",
"name": "Dog10K structural variants at NHEJ1",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "VcfTabixAdapter",
"uri": "dog10k_nhej1_svs.vcf.gz"
}
}jbrowse add-track dog10k_nhej1_svs.vcf.gz \
--trackId dog10k_nhej1_svs \
--name "Dog10K structural variants at NHEJ1" \
--assemblyNames UU_Cfam_GSD_1.0 \
--load copyIn JBrowse Desktop, or in any running JBrowse Web session, open a view on this track’s assembly, then File → Open track..., choose Add track from pasted JSON, and paste:
{
"type": "VariantTrack",
"trackId": "dog10k_nhej1_svs",
"name": "Dog10K structural variants at NHEJ1",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "VcfTabixAdapter",
"uri": "dog10k_nhej1_svs.vcf.gz"
}
}The sample rows keep the Dog10K IDs. layout renames them for the sidebar and
gives each group a swatch without touching the VCF. It is display state, the
same thing the tree sidebar writes when you rearrange rows by hand, so it
belongs on a session's track entry; a layout put in displays is ignored with
no error:
{
"defaultSession": {
"name": "NHEJ1 deletion",
"views": [
{
"type": "LinearGenomeView",
"assembly": "UU_Cfam_GSD_1.0",
"loc": "chr37:25,570,000-25,580,000",
"tracks": [
{
"trackId": "dog10k_nhej1_svs",
"type": "LinearMultiSampleVariantDisplay",
"layout": [
{
"name": "COLL000001",
"label": "Collie 1",
"color": "#0072B2"
},
{
"name": "CLUPGR000001",
"label": "Wolf 1",
"color": "#E69F00"
}
]
}
]
}
]
}
}jbrowse set-default-session --session - << 'EOF'
{
"name": "NHEJ1 deletion",
"views": [
{
"type": "LinearGenomeView",
"assembly": "UU_Cfam_GSD_1.0",
"loc": "chr37:25,570,000-25,580,000",
"tracks": [
{
"trackId": "dog10k_nhej1_svs",
"type": "LinearMultiSampleVariantDisplay",
"layout": [
{
"name": "COLL000001",
"label": "Collie 1",
"color": "#0072B2"
},
{
"name": "CLUPGR000001",
"label": "Wolf 1",
"color": "#E69F00"
}
]
}
]
}
]
}
EOFAdd the assembly's gene annotation above it, to read the deletion against NHEJ1's exons.
Reading the NHEJ1 deletion
The deletion falls inside an intron and clears no exon, which is how a variant this large can be common in a breed.
Checking the call against a curated source
The middle lane is OMIA, which curates the published causal variants of Mendelian traits in animals; its Collie eye anomaly record (OMIA 000218-9615) is this deletion. Its span was published on CanFam3.1 and lifted here with UCSC's chain, so the bar and the genotype column below it come from two publications by two routes:
curl -fO https://hgdownload.soe.ucsc.edu/goldenPath/canFam3/liftOver/canFam3ToCanFam4.over.chain.gz
./liftOver omia_canFam3.bed canFam3ToCanFam4.over.chain.gz lifted.bed unmapped.bed
wc -l < unmapped.bed # records the chain could not placeAn interval lifts as a unit, so a plain liftOver is enough. A BND carries its
partner coordinate inside ALT and needs more; see the
cancer SV tutorial.
{
"type": "FeatureTrack",
"trackId": "omia_dog_variants",
"name": "OMIA causal variants (dog)",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "Gff3TabixAdapter",
"uri": "omia_dog_variants.gff3.gz"
},
"displayDefaults": {
"labels": { "description": "jexl:feature.inheritance" }
}
}jbrowse add-track omia_dog_variants.gff3.gz \
--trackId omia_dog_variants \
--name "OMIA causal variants (dog)" \
--assemblyNames UU_Cfam_GSD_1.0 \
--displayDefaults '{"labels":{"description":"jexl:feature.inheritance"}}' \
--load copyIn JBrowse Desktop, or in any running JBrowse Web session, open a view on this track’s assembly, then File → Open track..., choose Add track from pasted JSON, and paste:
{
"type": "FeatureTrack",
"trackId": "omia_dog_variants",
"name": "OMIA causal variants (dog)",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "Gff3TabixAdapter",
"uri": "omia_dog_variants.gff3.gz"
},
"displayDefaults": {
"labels": { "description": "jexl:feature.inheritance" }
}
}The mode of inheritance is drawn as the feature's description. Click the bar for the rest of the record, including whether it reached canFam4 through a chain. A lifted record can be right about the locus and wrong about the base.
Filtering to one record
The window holds nine SV records, and the figure filters to this one:
{
"type": "VariantTrack",
"trackId": "dog10k_nhej1_svs",
"name": "Dog10K structural variants at NHEJ1",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "VcfTabixAdapter",
"uri": "dog10k_nhej1_svs.vcf.gz"
},
"displays": [
{
"type": "LinearMultiSampleVariantDisplay",
"displayId": "dog10k_nhej1_svs-LinearMultiSampleVariantDisplay",
"jexlFilters": ["feature.start == 25574004"]
}
]
}jbrowse add-track-json '{
"type": "VariantTrack",
"trackId": "dog10k_nhej1_svs",
"name": "Dog10K structural variants at NHEJ1",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "VcfTabixAdapter",
"uri": "dog10k_nhej1_svs.vcf.gz"
},
"displays": [
{
"type": "LinearMultiSampleVariantDisplay",
"displayId": "dog10k_nhej1_svs-LinearMultiSampleVariantDisplay",
"jexlFilters": ["feature.start == 25574004"]
}
]
}'In JBrowse Desktop, or in any running JBrowse Web session, open a view on this track’s assembly, then File → Open track..., choose Add track from pasted JSON, and paste:
{
"type": "VariantTrack",
"trackId": "dog10k_nhej1_svs",
"name": "Dog10K structural variants at NHEJ1",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "VcfTabixAdapter",
"uri": "dog10k_nhej1_svs.vcf.gz"
},
"displays": [
{
"type": "LinearMultiSampleVariantDisplay",
"displayId": "dog10k_nhej1_svs-LinearMultiSampleVariantDisplay",
"jexlFilters": ["feature.start == 25574004"]
}
]
}Unfiltered, a second deletion nested inside the 7.8 kb one paints yellow no-calls against the darkest blue and the two records read as one striped block. The nested deletion is missing in exactly the four dogs homozygous for the larger one, since a dog with no copy of the surrounding sequence has no reads to genotype it from:
# -i POS=… because -r is END-aware and would also return the deletion this one
# sits inside
bcftools query -r chr37:25578185-25578186 -i 'POS=25578185' \
-f '[%SAMPLE=%GT ]\n' dog10k_nhej1_svs.vcf.gz \
| tr ' ' '\n' | grep -v '=0/0'The Lancashire Heelers
Collie eye anomaly is reported in Lancashire Heelers, and none of the four sampled here carry the deletion. Four dogs is not a frequency estimate.
Two diet genes that run opposite ways
A 14.9 kb DUP at chr6:47,375,677 in the Michigan Manta callset spans the
pancreatic amylase gene end to end. Extra copies are the starch-digestion
signature of domestication
(Axelsson et al. 2013), and the record
separates dogs from wolves almost completely:
Breed_Dogs 1575 canids: 1568 hom alt, 6 hom ref, 1 het
Mixed/Other 12 canids: 12 hom alt
Village_Dogs 237 canids: 236 hom alt, 1 hom ref
Wolf 55 canids: 50 hom ref, 4 het, 1 hom altA 223 bp SINE insertion in pancreatic ribonuclease, chr15:18,164,072 in the Zenodo Paragraph set, runs the other way:
Breed_Dogs 1575 canids: 1574 hom ref, 1 het
Mixed/Other 12 canids: 12 hom ref
Village_Dogs 237 canids: 236 hom ref, 1 het
Wolf 55 canids: 29 hom ref, 26 hetOne panel is sliced from both callsets in the same order so the two lanes read row for row: two ordinary breeds, the three Arctic breeds, the two breeds holding a dog that departs from the amylase rule, the Alaskan village dogs, and every gray wolf labelled by country.
{
"type": "VariantTrack",
"trackId": "dog10k_amy2b_svs",
"name": "Dog10K structural variants at AMY2B (dogs and every wolf)",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "VcfTabixAdapter",
"uri": "dog10k_amy2b_svs.vcf.gz",
"samplesTsvLocation": { "uri": "dog10k_amy2b_samples.tsv" }
},
"displays": [
{
"type": "LinearMultiSampleVariantDisplay",
"colorBy": "group",
"height": 900
}
]
}jbrowse add-track-json '{
"type": "VariantTrack",
"trackId": "dog10k_amy2b_svs",
"name": "Dog10K structural variants at AMY2B (dogs and every wolf)",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "VcfTabixAdapter",
"uri": "dog10k_amy2b_svs.vcf.gz",
"samplesTsvLocation": { "uri": "dog10k_amy2b_samples.tsv" }
},
"displays": [
{
"type": "LinearMultiSampleVariantDisplay",
"colorBy": "group",
"height": 900
}
]
}'In JBrowse Desktop, or in any running JBrowse Web session, open a view on this track’s assembly, then File → Open track..., choose Add track from pasted JSON, and paste:
{
"type": "VariantTrack",
"trackId": "dog10k_amy2b_svs",
"name": "Dog10K structural variants at AMY2B (dogs and every wolf)",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "VcfTabixAdapter",
"uri": "dog10k_amy2b_svs.vcf.gz",
"samplesTsvLocation": { "uri": "dog10k_amy2b_samples.tsv" }
},
"displays": [
{
"type": "LinearMultiSampleVariantDisplay",
"colorBy": "group",
"height": 900
}
]
}Too many rows for a layout entry each, so the labels come from a TSV: first
column the sample name, every other column an attribute, and colorBy naming
the one that paints the swatch. The RNASE1 track is the same config with the
other slice's uri.
Two of the three Greenland Dogs lack the duplication; the third carries it, as does every Alaskan Malamute and Samoyed. The grey Czechoslovakian Wolfdog row is CZEC000003, the animal the local-ancestry tutorial paints wolf-derived blocks on. Every wolf carrying the insertion is heterozygous. Three of the six Iranian wolves carry the amylase duplication and none the ribonuclease insertion, while the Greek and Swedish wolves do the reverse.
A genotype column does not carry copy number: four copies and twenty are both
1/1. The CYP1A2 tutorial builds that measurement
from the SNV callset's per-sample DP, and dog10k_slc28a3_breed_cn and
dog10k_slc28a3_cohort_cn in this tutorial's config are the same pair of lanes
over a second duplication.
The FGF4 retrogene, read at its parent gene
The variant here is an insertion somewhere else in the genome, and what the callset holds at FGF4 is its footprint. Parker et al. (2009) tied breed-defining short legs to an expressed FGF4 retrogene, a processed copy of the transcript reinserted elsewhere. Made from spliced mRNA, it has no introns, so short reads from the retrocopy map to the parent's exons and stop at each splice site, and a short-read caller reads that pileup as a deletion of each intron. The callset cannot tell a retrocopy's footprint from a real deletion.
Parker et al. (2009) tied breed-defining short legs to an expressed FGF4 retrogene, a processed copy of the FGF4 transcript reinserted elsewhere. Processed means it was made from the spliced mRNA, so it has no introns: short reads from the retrocopy map to the parent's exons and stop at each splice site, and a short-read caller reads that pileup as a deletion of each intron.
Checking the records against the FGF4 introns
FGF4 has two introns, so a retrocopy should leave two records, each spanning
one intron end to end.
build_dog10k_fgf4_retrogene.sh
derives the introns from the RefSeq annotation and asserts each record against
them, with one base of slack at each breakpoint:
FGF4 RefSeq exons: 48869443-48869782, 48870315-48870418, 48870953-48873311
FGF4 RefSeq introns: 48869783-48870314 (532 bp), 48870419-48870952 (534 bp)
intron 48869783-48870314: called as a DEL of 532 bp at 48869783-48870314
intron 48870419-48870952: called as a DEL of 534 bp at 48870418-48870951Slicing the two records out
This locus comes from the Michigan aggregate Manta callset, which carries DUP
and INV records too. Selecting on POS keeps the two intron records:
SHARE=https://kiddlabshare.med.umich.edu/dog10K
SV=$SHARE/Manta-SV_2022-03-28/SV-genotype-v2.merge.agg_only.08032022.vcf.gz
bcftools view -r chr18:48865000-48876000 -S fgf4.samples --force-samples \
-i 'POS=48869782 || POS=48870417' \
-Oz -o dog10k_fgf4_svs.vcf.gz "$SV"
tabix -p vcf dog10k_fgf4_svs.vcf.gzfgf4.samples is whole breeds again: three breeds whose short legs are the
trait Parker et al. mapped, two spaniel breeds, two standard-proportioned
breeds, and the Greek gray wolves, labelled through a samples TSV with colorBy
on the breed group.
{
"type": "VariantTrack",
"trackId": "dog10k_fgf4_svs",
"name": "Dog10K structural variants at FGF4 (named breeds)",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "VcfTabixAdapter",
"uri": "dog10k_fgf4_svs.vcf.gz",
"samplesTsvLocation": { "uri": "dog10k_fgf4_samples.tsv" }
},
"displays": [
{
"type": "LinearMultiSampleVariantDisplay",
"colorBy": "group",
"height": 690
}
]
}jbrowse add-track-json '{
"type": "VariantTrack",
"trackId": "dog10k_fgf4_svs",
"name": "Dog10K structural variants at FGF4 (named breeds)",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "VcfTabixAdapter",
"uri": "dog10k_fgf4_svs.vcf.gz",
"samplesTsvLocation": { "uri": "dog10k_fgf4_samples.tsv" }
},
"displays": [
{
"type": "LinearMultiSampleVariantDisplay",
"colorBy": "group",
"height": 690
}
]
}'In JBrowse Desktop, or in any running JBrowse Web session, open a view on this track’s assembly, then File → Open track..., choose Add track from pasted JSON, and paste:
{
"type": "VariantTrack",
"trackId": "dog10k_fgf4_svs",
"name": "Dog10K structural variants at FGF4 (named breeds)",
"assemblyNames": ["UU_Cfam_GSD_1.0"],
"adapter": {
"type": "VcfTabixAdapter",
"uri": "dog10k_fgf4_svs.vcf.gz",
"samplesTsvLocation": { "uri": "dog10k_fgf4_samples.tsv" }
},
"displays": [
{
"type": "LinearMultiSampleVariantDisplay",
"colorBy": "group",
"height": 690
}
]
}Each record draws at its own coordinates, so the two blocks sit against the exons. Every carrier is heterozygous: the parent gene's introns are still on both chromosomes, so a carrier's pileup is always a mixture.
The two known FGF4 retrocopies
Two FGF4 retrocopies are known in dogs. Parker et al. tied one to short legs; Brown et al. (2017) tied a second, on a different chromosome, to chondrodystrophy and intervertebral disc disease, which is why breeds of ordinary proportions carry a copy too. Both leave the same footprint at the parent gene, so one record cannot say which. The spaniels are the rows where proportions and genotype disagree. Placing either insertion needs the other side of the junction, from a different callset.
The retrocopy itself, as sequence
Both copies were amplified, Sanger-sequenced and deposited, as MF040222 for the CFA18 insertion and MF040221 for the CFA12 one. One alignment per retrocopy, against the parent locus cut out as its own FASTA:
# `splice` chains across the parent's introns so both gaps land on the annotated
# ones; -c writes the base-level CIGAR the ribbons are drawn from. Rewrite the
# N operations it emits to D afterwards: those bases really are absent here.
samtools faidx parent.fa
minimap2 -x splice -c parent.fa FGF4retro-CFA12.fa > FGF4retro-CFA12.pafLoad each retrocopy as a one-contig assembly and its alignment as a
SyntenyTrack:
{
"type": "SyntenyTrack",
"trackId": "dog10k_fgf4_retro_cfa12",
"name": "FGF4 CFA12 retrocopy (MF040221) vs its parent gene",
"assemblyNames": ["FGF4retro-CFA12", "UU_Cfam_GSD_1.0"],
"adapter": {
"type": "PAFAdapter",
"uri": "dog10k_fgf4_retro_cfa12.paf",
"queryAssembly": "FGF4retro-CFA12",
"targetAssembly": "UU_Cfam_GSD_1.0"
}
}jbrowse add-track-json '{
"type": "SyntenyTrack",
"trackId": "dog10k_fgf4_retro_cfa12",
"name": "FGF4 CFA12 retrocopy (MF040221) vs its parent gene",
"assemblyNames": ["FGF4retro-CFA12", "UU_Cfam_GSD_1.0"],
"adapter": {
"type": "PAFAdapter",
"uri": "dog10k_fgf4_retro_cfa12.paf",
"queryAssembly": "FGF4retro-CFA12",
"targetAssembly": "UU_Cfam_GSD_1.0"
}
}'In JBrowse Desktop, or in any running JBrowse Web session, open a view on this track’s assembly, then File → Open track..., choose Add track from pasted JSON, and paste:
{
"type": "SyntenyTrack",
"trackId": "dog10k_fgf4_retro_cfa12",
"name": "FGF4 CFA12 retrocopy (MF040221) vs its parent gene",
"assemblyNames": ["FGF4retro-CFA12", "UU_Cfam_GSD_1.0"],
"adapter": {
"type": "PAFAdapter",
"uri": "dog10k_fgf4_retro_cfa12.paf",
"queryAssembly": "FGF4retro-CFA12",
"targetAssembly": "UU_Cfam_GSD_1.0"
}
}assemblyNames is ordered [query, target], the reverse of minimap2's argument
order. Each GenBank record carries a feature table, which the build script
writes out as GFF3 for the retrocopy row's gene model; it requires the CDS to be
a single interval, where the parent's is three.
Put the parent gene between the two retrocopies. Both align to the same three exons, so each intron is one gap seen twice. The window stops where the CFA18 alignment does, so the CFA12 ribbon runs on past it.
Set the synteny view's indel drawing to Transparent indels. Colored indels name each CIGAR operation from the side it is read, so one gap is a deletion above the parent row and an insertion below it.
The two records agree at 207 codons but differ in how much UTR they took. Neither places the insertion: the deposited sequence ends at the poly(A) tail.
Genotypes across the collection
The same two records genotyped over every canid the callset carries, printed by the build script:
Genotype counts per group, at the intron 1 record (chr18:48869782):
Breed_Dogs 1575 canids: 1177 hom ref, 381 het, 12 no call, 5 hom alt
Mixed/Other 12 canids: 10 hom ref, 2 het
Village_Dogs 237 canids: 198 hom ref, 39 het
Wolf 55 canids: 55 hom ref
of 290 breeds with two or more animals: 52 carry it in every animal, 198 in none
1831 of 1879 canids get the same call from both: 97.4%
most common (intron 1, intron 2) pairs: (0/0, 0/0) x1422 (0/1, 0/1) x409No wolf carries it. Manta called the two introns independently, so the same
animals landing on both records is a check. The whole-collection track is in the
config as dog10k_fgf4_cohort_svs; 1,879 rows in a few hundred pixels puts each
row well under a pixel, where rows alias.
Where to go next
Schall and Kidd's table of clade-associated SVs is the place to pick the next locus. For the retrogene shape, any gene whose introns are all called deleted in some animals and not others is a candidate, and the check is the one the script runs: do the records match the annotated introns to the base.
Reproduce it end to end
build_dog10k_nhej1_sv.sh
builds the track:
curl -fO https://raw.githubusercontent.com/GMOD/jbrowse-components/main/scripts/build_dog10k_nhej1_sv.sh
bash build_dog10k_nhej1_sv.sh # writes ./dog10k_sv_build/It downloads the Dog10K sample table, derives the breed lists from it, slices the locus out of the Zenodo genotype VCF, and prints the deletion's genotypes.
build_omia_dog_variants.sh
builds the OMIA lane:
curl -fO https://raw.githubusercontent.com/GMOD/jbrowse-components/main/scripts/build_omia_dog_variants.sh
bash build_omia_dog_variants.sh # writes ./omia_dog_build/OMIA publishes no coordinate API, so this reads the nightly mysqldump, keeps the dog records, lifts the CanFam3.1 majority with UCSC's chain, and prints what each assembly contributed and how many the lift dropped. The database is curated continuously, so another day gives a different count.
build_dog10k_amy2b_sv.sh
builds the amylase track:
curl -fO https://raw.githubusercontent.com/GMOD/jbrowse-components/main/scripts/build_dog10k_amy2b_sv.sh
bash build_dog10k_amy2b_sv.sh # writes ./dog10k_amy2b_build/It derives the panel and the label TSV from the sample table, slices the duplication record out of the Manta callset, then genotypes it over every canid: the tally quoted above, the non-carrier dogs and carrier wolves by name, and the wolves by country.
build_dog10k_slc28a3_cn.sh
builds the copy-number tracks the same way:
curl -fO https://raw.githubusercontent.com/GMOD/jbrowse-components/main/scripts/build_dog10k_slc28a3_cn.sh
bash build_dog10k_slc28a3_cn.sh # writes ./dog10k_slc28a3_cn_build/It prints each panel animal's copy number over the duplication. Its first route
needs only bcftools; the second re-measures six animals from their SRA runs,
which needs an aligner and about 35 GB of scratch.
Two more build the FGF4 locus:
BASE=https://raw.githubusercontent.com/GMOD/jbrowse-components/main/scripts
curl -fO $BASE/build_dog10k_fgf4_retrogene.sh
curl -fO $BASE/build_dog10k_fgf4_synteny.sh
bash build_dog10k_fgf4_retrogene.sh # writes ./dog10k_fgf4_build/
bash build_dog10k_fgf4_synteny.sh # writes ./dog10k_fgf4_synteny_build/build_dog10k_fgf4_retrogene.sh
derives the panel and the label TSV from the sample table, checks both records
against the RefSeq introns, slices them out of the callset for the panel and for
the whole collection, and prints the genotype counts quoted above.
build_dog10k_fgf4_synteny.sh
fetches both GenBank records and the parent locus, writes each record's feature
table out as GFF3, aligns each retrocopy, and rewrites the PAF into absolute
chr18 coordinates. It exits non-zero unless every gap in both alignments lands
on an annotated FGF4 intron and each deposited CDS is a single interval.
See also
- A loss-of-function allele across breeds (Dog10K)
- A selected haplotype (Dog10K)
- Local ancestry (Dog10K)
- Structural variants (1000 Genomes)
- CNV across a population (1000 Genomes)
- Multi-sample variant display
- Config guide: Variant track
- SV visualization
- Linear synteny view
References
- Axelsson et al. (2013). The genomic signature of dog domestication reveals adaptation to a starch-rich diet
- Schall & Kidd (2025). Integrative genotyping and analysis of canine structural variation using long-read and short-read data
- Parker et al. (2007). Breed relationships facilitate fine-mapping studies: a 7.8-kb deletion cosegregates with Collie eye anomaly across multiple dog breeds
- Parker et al. (2009). An expressed fgf4 retrogene is associated with breed-defining chondrodysplasia in domestic dogs
- Brown et al. (2017). FGF4 retrogene on CFA12 is responsible for chondrodystrophy and intervertebral disc disease in dogs
- Meadows et al. (2023). Genome sequencing of 2000 canids by the Dog10K consortium advances the understanding of demography, genome function and architecture
Feedback on this tutorial is welcome: contact us.