Oncofusion

start / graph / absence-q130-the-index-fusion-version-of-this-test-is-fully-deposited-costs-1-99-gb-and-has-never-been-run

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q130-does-the-eviction-gradient-exist-in-h3k79me2

RS4;11 - the human KMT2A::AFF1 index-fusion line - has a deposited H3K79me2 pair, a MENIN pair, an MLL1-N pair, a DOT1L pair and an input under the same menin inhibitor in GSE127507. Every track needed to run this beat's test on the index fusion exists. The download is 1.99 GB measured by HEAD request, and nobody has run it.

also called the index-fusion gap · RS4;11 has all of it · 1.99 GB and one binner

The disease this whole project is about is driven by a specific fusion, and a human cell line carrying it has all the measurements needed to test this beat's finding - already public since 2019. Running it needs about two gigabytes of downloads and a piece of code that was deleted.

The same thing, for a scientist

**Everything is deposited.** GSE127507's RS4;11 arm, read from the series' own 26 sample records [E1]: | sample | GSM | Content-Length | |---|---|---:| | `RS411_ChIP_H3K79me2-CST_DMSO_d3` | GSM3635697 | 313,943,028 | | `RS411_ChIP_H3K79me2-CST_VTP_330nM_d3` | GSM3635698 | 329,755,483 | | `RS411_ChIP_MENIN_DMSO_d3` | GSM3635693 | 176,146,076 | | `RS411_ChIP_MENIN_VTP_d3` | GSM3635694 | 165,977,893 | | `RS411_ChIP_MLL1n_DMSO_d3` | GSM3635695 | 191,650,790 | | `RS411_ChIP_MLL1n_VTP_d3` | GSM3635696 | 176,999,077 | | `RS411_ChIP_DOT1L_DMSO_d3` | GSM3635691 | 107,863,985 | | `RS411_ChIP_DOT1L_VTP_d3` | GSM3635692 | 164,586,145 | | `RS411_ChIP_input_DMSO_d3` | GSM3635699 | 362,257,987 | | **total** | | **1,989,180,464 bytes = 1.99 GB** | **Every exposure is d3**, so unlike MOLM13's MENIN pair there is no exposure mismatch. **The H3K79me2 antibody is named** (`H3K79me2-CST`). **RS4;11 is KMT2A::AFF1** - this graph's index fusion and its index lineage [E1]. **What is missing is not data. It is two things that were thrown away and one that was never written** [E2]: 1. **`C:/Temp/q118` has been DELETED**, and with it `q118_hg19_blacklist.bed`, `q118_hg19_chroms.json` and the `Q118Bin` class - the human binner. `Q117Bin` iterates a hardcoded mm10 chr1-19+X list and cannot be used on a human file. So a human binner has to be rebuilt or recovered before anything can be binned. 2. **Q118 binned only RS4;11's MLL1-N, MENIN and input** (GSM3635693-3635696, GSM3635699). It never binned RS4;11's H3K79me2 or DOT1L, and its binned TSVs are gone with the directory. 3. **Nobody has asked this question of RS4;11.** Q118 measured retained occupancy at co-occupied sites; the eviction-depth stratification did not exist until Q117 defined it in cycle 68, and it has never been applied to a human KMT2A::AFF1 cell. **`absence_class` is `measured-not-analysed`** - the assay was run, the data is public, and the analysis has never been done. That is the cheapest class of absence in this corpus's own taxonomy: compute, no wet work [E2].

Go deeper
What running it would settle, and what it would not

Would settle: whether the eviction gradient exists in the index fusion and the index lineage. Every measurement in this beat is KMT2A::MLLT3 - murine MLL-AF9 and human MOLM13 - and this corpus's commonest documented synthesis failure is generalising exactly that to KMT2A::AFF1.

Would also settle whether DOT1L occupancy is graded in the index fusion, because RS4;11's DOT1L pair is deposited too, and the AF4/AFF1 axis is the one that recruits DOT1L in this fusion.

Would NOT settle the drug-specificity half: GSE127507 has no SGC0946 arm in any line, so the flat-DOT1L-inhibitor control cannot be reproduced there. It would also inherit the shared-epitope problem in its stratifier - MLL1-N sees wild-type KMT2A too - although RS4;11's MENIN pair gives a second stratifier at a matched d3 exposure, which MOLM13's did not.

The scratch-directory pattern, because it will recur

Of the beats whose scratch directories once existed, C:/Temp/q118 is gone while q117, q120, q122, q124, q125, q126 and q127 survive. What was lost was not the expensive part. Q120's binned human TSVs survive in its own directory and were the basis of this beat's human result - but they were nearly unusable, because the universe definition lived in q118 and the hg19 blacklist had to be re-fetched.

The general rule this suggests: a beat's binned outputs are worthless without the blacklist, the chromosome sizes and the binner that defined them, and those three files total under 200 kB. They should be copied into every scratch directory that uses them, not referenced across directories. Q124 did this correctly - q124_mm10_blacklist.bed and q124_mm10_chroms.json are in its own directory - and that is why the murine arm of this beat cost nothing.

Facts and figures
absence classmeasured-not-analysed
absence class noteEvery RS4;11 track needed is public since 2019 in GSE127507. The analysis needs 1.99 GB of download, a rebuilt human binner and about an hour of compute. No wet work, no new cohort, no new reagent.
absence class assigned2026-09-23 (Q130)
total download bytes1989180464
measured howHTTP HEAD Content-Length on each supplementary file, 2026-09-23
what is absentan eviction-depth-stratified H3K79me2 measurement under a menin inhibitor in any KMT2A::AFF1 cell, human or murine
epigenetic readoutH3K79me2 under a menin inhibitor in the index fusion - not yet measured here
accessions
[
  "GSM3635691",
  "GSM3635692",
  "GSM3635693",
  "GSM3635694",
  "GSM3635695",
  "GSM3635696",
  "GSM3635697",
  "GSM3635698",
  "GSM3635699"
]
missing infrastructure
[
  "C:/Temp/q118 deleted: q118_hg19_blacklist.bed, q118_hg19_chroms.json, Q118Bin",
  "Q117Bin hardcodes mm10 chr1-19+X and cannot bin a human file"
]
recoverable substitutes
[
  "hg19 chromosome sizes: C:/Temp/q125/q125_hg19_chroms.json",
  "hg19 ENCODE blacklist v2: re-fetched by Q130, 7,731 bytes, in C:/Temp/q130/q130_hg19_blacklist.bed"
]
Still open
  1. Is the 1.99 GB worth it? It is the only route from 'the gradient exists in KMT2A::MLLT3' to 'the gradient exists in the disease this graph is about', and no cheaper route has been identified.
  2. Can Q118Bin be recovered from the repository rather than rebuilt? Its source was in C:/Temp/q118 and no copy has been found in the repo; Q117Bin.java survives in C:/Temp/q117 and differs only in its hardcoded chromosome list, so a rebuild is small.
  3. Which other beats' binned outputs are orphaned from their universe definitions? A sweep of the surviving scratch directories for a self-contained blacklist and chrom file would price the risk once.
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    GSE127507 deposits, for RS4;11 (human KMT2A::AFF1), an H3K79me2 DMSO/VTP pair (GSM3635697/698), a MENIN pair (GSM3635693/694), an MLL1-N pair (GSM3635695/696), a DOT1L pair (GSM3635691/692) and an input (GSM3635699), all at d3 exposure, totalling 1,989,180,464 bytes measured by HEAD request on 2026-09-23.

    What this evidence cannot show

    Sizes are Content-Length headers from the NCBI FTP-over-HTTPS endpoint, read this session; they are the compressed TDF sizes, not the binned output sizes. The RS4;11 arm has NO SGC0946 sample and no second input, so a DMSO-versus-DMSO selection control and a drug-shape control are both unavailable in it. One replicate per antibody per condition.

    Evidence that argues against this (1)
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    The index-fusion analysis has never been run and the obstacle is infrastructure rather than data: C:/Temp/q118 has been deleted along with the hg19 blacklist, the hg19 chromosome sizes and the Q118Bin human binner, and Q118 in any case binned only RS4;11's MLL1-N, MENIN and input, never its H3K79me2 or DOT1L.

    What this evidence cannot show

    'Deleted' is inferred from C:/Temp/q118 being an empty or absent directory on 2026-09-23 while q117, q120, q122, q124, q125, q126 and q127 survive; it is not known who deleted it or when. Q118's committed results.json in the repository survives and records which samples it binned, which is how the second half of the claim is established.

    Evidence that argues against this (1)
    • Calling this measured-not-analysed assumes nobody outside this project has run it; the founding paper's authors had the same data. Krivtsov AV et al. 2019
Connected to