genomes

Genome-level annotation used by sesame-cli (CNV binning, ideogram plotting, and region/gene tracks) and other zhou-lab plotting tools. Kept in its own repo, separate from the platform/probe annotation in InfiniumAnnotation, so it can be reused independently of any array platform.

Layout

One folder per genome build, versioned by git tag (v1, v2, …):

<build>/SHA256SUMS               coreutils digests of the files below (the trust anchor)
<build>/seqinfo.tsv.gz           chrom <TAB> length              (genome tiling)
<build>/gaps.tsv.gz              chrom <TAB> start <TAB> end     (assembly gaps to skip)
<build>/cytoband.tsv.gz          chrom start end band stain      (ideogram, plot-only)
<build>/genes.bed.gz (+.tbi)     GENCODE transcript models (tabix BED12+)

seqinfo + gaps drive getBinCoordinates (tile the genome into bins, carve out the gaps); cytoband is only for the ideogram; these three are exported from sesameData by tools/export_genomeinfo.R in the sesame-cli repo. genes.* is built here from GENCODE by tools/build_genes.sh (see Gene models below).

Gene models

genes.bed.gz is a tabix-indexed transcript-model table — the plot-ready gene track for sesame region and cinderplot's locus browser. One row per transcript, sorted, bgzipped, and indexed (tabix -p bed), so a locus is a single random-access query:

tabix hg38/genes.bed.gz chr20:44616522-44655233

It is BED12 + five named columns (tab-separated; no trailing comma in the block lists):

1 chrom          6 strand         11 blockSizes      16 transcript_name
2 chromStart(0)  7 thickStart     12 blockStarts     17 transcript_type
3 chromEnd       8 thickEnd       13 gene_id
4 transcript_id  9 itemRgb (0)    14 gene_name
5 score (0)     10 blockCount     15 gene_type

thickStart/thickEnd are the CDS span ([min CDS start, max CDS end]); a non-coding transcript has thickStart == thickEnd == chromEnd (no thick drawn). This reproduces the semantics of sesameData:::build_GENCODE_gtf (which built genomeInfo$txns) — validated exon-for-exon against the ADA locus — but as a plain indexable BED, not an R GRangesList.

Each build uses the GENCODE release its sesameData txns came from. The file name is unversioned (like seqinfo/gaps/cytoband) — the release is pinned by git tag, and the exact source URL is recorded next to the file in genes.bed.gz.source. Bumping GENCODE means rerunning the builder and cutting a new tag.

build	GENCODE release	source GTF
hg38	v36	gencode.v36.annotation.gtf.gz
mm10	vM25	gencode.vM25.annotation.gtf.gz
mm39	vM31	gencode.vM31.annotation.gtf.gz

# tools/build_genes.sh <build> <gencode_gtf_url> [outroot]   (needs bgzip/tabix)
cd tools && ./build_genes.sh hg38 \
  https://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_human/release_36/gencode.v36.annotation.gtf.gz ..

Fetch

sesame fetch genome hg38     # -> <store>/genome/hg38/{seqinfo,gaps,cytoband,genes.*}

sesame pulls <build>/SHA256SUMS first, verifies it against a digest compiled into the build, then verifies every file it lists against it — a hard chain, and the file list is exactly SHA256SUMS, so the gene models come along automatically. Files are raw-served over https://github.com/zhou-lab/genomes/raw/<tag>/<build>/<file>.

Builds

build	seqinfo / gaps / cytoband	gene models
hg38	sesameData_getGenomeInfo("hg38")	GENCODE v36
mm10	sesameData_getGenomeInfo("mm10")	GENCODE vM25
mm39	sesameData_getGenomeInfo("mm39")	GENCODE vM31