# Structural variants (1000 Genomes) **TL;DR:** take one whole-gene deletion out of the 1000 Genomes ensemble SV callset, sort the cohort's genotypes at it, then open three of those samples' reads and watch the coverage go to zero, halve, and stay flat. The page ends on a call from the same cohort whose coverage does none of that. ## Prerequisites - nothing to install: the demo instance already carries the callset and the alignments ## Where the data comes from The 1000 Genomes 2022 high-coverage ensemble SV callset ([Byrska-Bishop et al., 2022](https://doi.org/10.1016/j.cell.2022.08.004)), read against three of its own samples' CRAMs and QuicK-mer2 copy number for the whole cohort. - the ensemble SV callset, 3202 samples. EBI publishes it and nobody mirrors it, so the demo reads our own byte-for-byte copy[^ebi]: https://jbrowse.org/demos/1000g/1KGP_3202.Illumina_ensemble_callset.freeze_V1.vcf.gz - HG00113 (homozygous alt) high-coverage CRAM: https://1000genomes.s3.amazonaws.com/1000G_2504_high_coverage/data/ERR3240129/HG00113.final.cram - HG00096 (heterozygous) high-coverage CRAM: https://1000genomes.s3.amazonaws.com/1000G_2504_high_coverage/data/ERR3240114/HG00096.final.cram - HG00097 (homozygous reference) high-coverage CRAM: https://1000genomes.s3.amazonaws.com/1000G_2504_high_coverage/data/ERR3240115/HG00097.final.cram - QuicK-mer2 copy number for the cohort, the store the [copy-number tutorial](https://jbrowse.org/jb2/docs/tutorials/population_cnv) also reads. A directory of chunks rather than a file, so it is the `uri` an adapter takes: https://jbrowse.org/demos/1000g/qm2_cn_1kb.zarr ## The 1000 Genomes SV callset The [1000 Genomes Project](https://www.internationalgenome.org/) sequenced 2,504 individuals across 26 populations. The 2022 high-coverage re-analysis ([Byrska-Bishop et al., 2022](https://doi.org/10.1016/j.cell.2022.08.004)) called deletions, insertions, inversions and translocations with per-sample genotypes across all 3,202 individuals. `HGSV_1821` is a deletion on chr1 spanning the whole of _RHD_, so samples called homozygous carry no copy of that gene. Deleting _RHD_ is the most common cause of the RhD-negative blood type. The call is `PASS` and common enough to fill all three genotype classes, and read depth settles whether a gene is present twice, once, or not at all. The tracks are added with `jbrowse add-track`. The callset is bgzip-compressed and tabix-indexed, and the alignment tracks stream the CRAMs from the urls above. In the track selector, enable the 1KGP 2022 Illumina ensemble SV callset under **1000 Genomes → SV callsets**, listed by its file name **1KGP_3202.Illumina_ensemble_callset.freeze_V1.vcf**, then navigate to `chr1:25,200,000-25,400,000`, the window the figures below use. Everything below works the same in [JBrowse Desktop](https://jbrowse.org/jb2/docs/quickstart_desktop), which opens your own VCF and BAM files from disk. ## Genotypes across the cohort Switch the track to the **Multi-sample variant display (regular)** from the track menu. Each sample becomes a row drawn at the variant's genomic span, so the deletion is a wide block. Clicking it opens the feature details panel, whose **SAMPLES** section lists every sample's genotype, read depth and other per-sample fields. Rows arrive in the callset's order. Right-click the deletion and pick **Sort by genotype** to order rows by genotype at that call, then by how far each keeps matching its neighbours. **Clustering → Cluster rows by genotype...** in the track menu keys every row on the whole window and draws the dendrogram. Dark blue is no copy of _RHD_, light blue one, grey two, and the olive stripe is a separate nested call. A matrix cell says a sample carries something at that column, not which call. Loading the same VCF again in the ordinary variant display puts each record on its own row with its id, class and size. Cell coloring by **SV type** is the other way to ask, shown in the [multi-variant track guide](https://jbrowse.org/jb2/docs/user_guides/multivariant_track). Three lanes read below, over NCBI RefSeq genes: - the callset as a genotype matrix, one row per sample, sorted by genotype at the _RHD_ deletion - the same records drawn ordinarily and colored by SV class, so a band in the matrix reads off a named record - QuicK-mer2 copy number for 2504 individuals, one row each and clustered on this window, where blue is a copy lost against the diploid white and red a copy gained The olive no-call column is a copy-number gain in the lane beneath it.