Oncofusion

start / graph / ana-q118-the-genome-build-is-hg19-verified-from-nine-files-own-chromosome-extents

Analysisestablished confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.Q118-menin-at-the-human-index-fusion

The genome build of GSE127507 is hg19, verified from the chromosome extents written inside each of the nine downloaded .tdf files - 9 bp from hg19 and 2,818,723 bp from hg38 on chr1, in every file - and independently from the deposited TSS table's own coordinates

also called hg19 verification, GSE127507 · build checked from the file and not the metadata

Before measuring anything, the reference genome each data file was aligned to was read out of the files themselves rather than taken from the deposit's description. All nine agree it is the 2009 human reference, and so does the separate processed table.

The same thing, for a scientist

A corpus entry in this project asserted a genome build wrongly two cycles ago, in both directions, so the build here is established from the data and not from `!Sample_data_processing = Genome_build: GRCh37/hg19`. **Route 1 - the `.tdf` files' own internal extents.** Each TDF stores a per-chromosome `tileWidth` at `/<chr>/z0/mean` which is the declared chromosome length. Compared against live UCSC chromosome sizes for hg19 and hg38 fetched the same day, for all **nine** downloaded files [E1]: | | value | |---|---:| | declared chr1 length, all nine files | **249,250,624** | | hg19 chr1 | 249,250,621 | | hg38 chr1 | 248,956,422 | | max absolute difference vs hg19, over chr1-22 + chrX, worst file | **9 bp** | | max absolute difference vs hg38, same | **2,818,723 bp** | | NaN entries in any file | **0** | A 313,000-fold discrimination. The 9 bp offset is the TDF tiling rounding, not a build difference. **The verdict does not depend on the `genomeId` header field**, which also reads hg19. **Route 2 - the deposited ROSE TSS table's own coordinates**, independent of route 1 [E2]. Maximum `TSS_region_end` per chromosome over 21,461 windows on 23 chromosomes: **0 chromosomes exceed their hg19 length and 13 exceed their hg38 length** (chr1 reaches 249,203,441, which is inside hg19's 249,250,621 and outside hg38's 248,956,422). Two independent routes, same answer. Every coordinate-joined annotation in this beat - the ENCODE blacklist v2 and `cpgIslandExt` - was therefore taken in hg19 and no liftover was performed anywhere.

Facts and figures
scriptanalyses/q118-menin-at-the-human-index-fusion/binsB.py (genome_build_verification block)
statusrun
result summaryhg19 confirmed by two independent routes: the nine TDF files' own declared chromosome extents (9 bp from hg19, 2,818,723 bp from hg38) and the deposited TSS table's own coordinates (0 of 23 chromosomes exceed hg19, 13 of 23 exceed hg38).
genome buildhg19 / GRCh37
n files verified9
n files in series26
inputs
[
  "nine GSE127507 .deduped.tdf files",
  "UCSC hg19 and hg38 chromosome sizes"
]
Still open
  1. The other 17 deposited .tdf files were not opened. A series that mixed builds between arms would be invisible to this check, and Q117 found a series that mixed PROCESSING batches with a twenty-fold difference in artefact load, signalled only by the filename - so the possibility is not theoretical.
  2. hg18 was not tested as a candidate build. It is excluded by the chr1 declared length (hg18 chr1 is 247,249,719) but that comparison was not run programmatically.
Every source behind this entry (2)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    All nine downloaded GSE127507 .tdf files declare chr1 as 249,250,624 bp, within 9 bp of hg19 on every one of chr1-chr22 and chrX and 2,818,723 bp from hg38 on chr1, with zero NaN coverage entries.

    What this evidence cannot show

    Nine of the 26 deposited files were downloaded (RS4;11 MLL1n x2, MENIN x2, input x1; MOLM13 MLL1n x2, input x2). The other 17 were not opened, so this verifies the build of the files this beat measured and not of the series. A declared chromosome length is written by the tool that made the file and is strong evidence of the reference used, but it is not the alignment itself.

    Evidence that argues against this (1)
    • A chromosome-length match cannot detect a within-chromosome coordinate error such as a bad liftover that preserves length. Nothing here excludes that; what makes it unlikely is that the top-ranked bins land on canonical KMT2A-r target promoters (JMJD1C, CDK6, MYB, RUNX2, IGF2BP2), which a shifted coordinate system would not produce. Iqbal Lab oncofusion knowledge graph 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The deposited ROSE TSS signal table's own coordinates place zero of 23 chromosomes beyond their hg19 length and 13 of 23 beyond their hg38 length, with chr1 reaching 249,203,441.

    What this evidence cannot show

    A coordinate exceeding a build's chromosome length is proof against that build; a coordinate inside it is only consistent with the build. Thirteen chromosomes exceeding hg38 is decisive against hg38; hg19 is then the build the metadata claims and the one everything else is consistent with. hg18 was not tested.

    Evidence that argues against this (1)
    • The TSS table's coordinates are the depositors' ROSE output, not the alignment, so they inherit whatever build the depositors' RefGene annotation used rather than proving the build of the reads. Route 1 - the extents written inside the .tdf files by igvtools at the time the coverage was made - is the stronger of the two, and this route is a consistency check on it rather than an independent proof. NCBI Gene Expression Omnibus 2019
Connected to