Oncofusion

start / graph / ana-gse299100-supplies-the-missing-selection-control-and-it-is-in-kmt2a-aff1-itself

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.

GSE299100 supplies the vehicle-versus-vehicle pair that neither GSE260742 nor GSE127507 could - two independent control hairpins, both DMSO, same date, uniformly processed to hg38_rmdup.tdf, for MEN1, MLL1 and input - in MV4-11, which Cellosaurus confirms carries KMT2A::AFF1 though it presents as monocytic rather than the infant B-ALL this project studies. It also holds MEN1 and MLL1 ChIP under five different menin inhibitors in that line, and the six control files total 1.87 GB

also called GSE299100 has the missing control · the vehicle pair exists and it is in KMT2A::AFF1 · shLuc-1 versus shLuc-2

Four recent results depend on sorting genomic regions by how much a drug removes a protein, then measuring a second change in the same regions - a design that can invent a pattern from nothing. Checking it needs two untreated samples that differ only by chance, and the two datasets this project had been using could not provide such a pair. A search of public deposits found one that can, and it is better than expected: the samples come from the exact leukaemia subtype this project studies rather than a mouse model of a related one, and the same deposit also compares five different drugs of the class in that cell line.

The same thing, for a scientist

**Statement.** Cycle 99 established that the selection control has nowhere clean to run: **GSE260742** has within-condition replicates but **mixed processing** (`filt.bam.tdf` vs `tdf`), and **GSE127507** is uniformly `.deduped.tdf` but has **no replicates at all**. Finding a third deposit was named the single blocking task for the thread. A GEO search over menin-inhibitor chromatin series (112 hits) found one. **GSE299100** — *UBR5 Mediates Menin Inhibitor Dependent Menin/KMT2A Complex Stability [ChIP-seq]*, 64 samples — supplies the pair, in **MV4-11**. **MV4-11 carries t(4;11), KMT2A::AFF1.** Every chromatin result this thread rests on comes from murine MLL-AF9 (KMT2A::MLLT3) or human MOLM-13 (KMT2A::MLLT3). **This is the first deposit reached in this thread that is in the fusion the project exists for.** **The vehicle-versus-vehicle pair, three times over.** Two independent control hairpins, `shLuc-1` and `shLuc-2`, both DMSO 72 h, **both sequenced on 230927**, both deposited as **`_hg38_rmdup.tdf`**: | assay | shLuc-1 | shLuc-2 | |---|---|---| | **MEN1** | `GSM9032979` | `GSM9032985` | | **MLL1** | `GSM9032981` | `GSM9032987` | | **input** | `GSM9032977` | `GSM9032983` | **Two non-targeting hairpins in the same line, same treatment, same day, same processing** — biologically equivalent, so any quartile gradient between them is manufactured by the method. That is precisely the control cycle 83 first named and that Q125, Q130 and Q131 could not run. **And the same deposit carries a five-drug comparison in KMT2A::AFF1.** `MV411_ChIP` MEN1 and MLL1 at 230913, all `_hg38_rmdup.tdf`, under **DMSO, DS-1594b, JNJ, KO-539 and SNDX-5613** at 50 nM 72 h, with matched inputs for each. This project has repeatedly treated "menin inhibitor" as one thing; **five of them side by side in the index fusion is a direct test of whether it is.** **Nothing here is computed yet.** This is a deposit survey from GEO sample titles and supplementary filenames. What it establishes is that **the blocking task is cleared** and what the next beat should do. **BOTH CHECKS THIS ENTRY NAMED WERE RUN BEFORE IT SHIPPED.** **MV4-11's fusion is verified, not assumed.** Cellosaurus `CVCL_0064` records *"Sequence variation: Gene fusion; AFF1 + KMT2A; Name(s)=KMT2A-AFF1, MLL-AFF1, ALL-1/AF4"* citing PubMed 1423625 and 8358709. **But the same record adds two things that qualify the claim:** the line also carries a **FLT3 internal tandem duplication**, and its disease annotation is **childhood acute *monocytic* leukaemia**, not B-lymphoblastic. **So MV4-11 is the right fusion in the wrong lineage presentation** — closer to the PI's subject than murine MLL-AF9 or MOLM-13, and not the infant B-ALL itself. **The download is affordable.** The six vehicle-pair files total **1,865,230,942 bytes ≈ 1.87 GB**, individually 271–345 MB. Taken one at a time, binned and deleted — the pattern Q125 used for 5.33 GB with nothing retained — **peak disk use is about 345 MB**, which C: absorbs comfortably at 15.2 GB free.

Go deeper
What still has to be checked before anyone trusts this

The processing claim rests on filenames, not on the depositor's methods. Every sample above ends _hg38_rmdup.tdf, which is a strong signal of one pipeline — and it is exactly the kind of inference that went wrong in GSE260742, where filt.bam.tdf versus tdf was visible in the filenames and nobody looked until a control depended on it. Here I did look, and the filenames agree; the methods section has not been read.

shLuc-1 and shLuc-2 are two hairpins, not two aliquots. They are independent lentiviral integrations into the same line, so they differ by clonal selection and by whatever off-target effects a non-targeting hairpin has. That makes the pair a conservative control — it contains real biological variation that a technical replicate would not — which is the right direction for a falsification test but means a positive result would need care in interpretation.

The stratifier has to be rebuilt for this deposit. Q117's eviction quartiles are defined on FLAG occupancy in a murine construct and do not transfer. In GSE299100 the natural stratifier is MEN1 or MLL1 loss between DMSO and drug, which is closer to the biology the quartiles were meant to capture than a FLAG tag on an over-expressed construct — but it is a different instrument and its own coverage gradient would need measuring, since cycle 99 found the murine quartiles carry a 15% input-coverage gradient at Spearman −1.000.

The download has now been measured: 1.87 GB for the six vehicle-pair files, 271–345 MB each. That is the same order as the RS4;11 arm deferred at 1.99 GB — but taken one at a time with bin-and-delete the peak is ~345 MB, which is affordable on C: at 15.2 GB even with D: at 9.7 GB. The RS4;11 deferral should probably be revisited on the same reasoning.

Facts and figures
questionis there a deposit where the vehicle-versus-vehicle selection control CAN be run?
answerYES - GSE299100, and it is in MV4-11 which carries KMT2A::AFF1
what is NOT donenothing is computed. No download, no bin, no genome-build verification beyond the hg38 string, no methods section read, no download size measured.
the stratifier must be rebuiltQ117's eviction quartiles are defined on FLAG occupancy in a murine construct and do not transfer. MEN1 or MLL1 loss between DMSO and drug is the natural substitute here, and its own coverage gradient would need measuring - cycle 99 found the murine quartiles carry a 15% input-coverage gradient at Spearman -1.000.
absence classnever-measured
absence class noteNo vehicle-versus-vehicle selection control has ever been run for this design in any deposit. This entry identifies where it can be, and does not run it.
series
{
  "accession": "GSE299100",
  "n_samples": 64,
  "title": "UBR5 Mediates Menin Inhibitor Dependent Menin/KMT2A Complex Stability [ChIP-seq]",
  "cell_line": "MV4-11 and derivatives",
  "fusion": "t(4;11) KMT2A::AFF1"
}
the vehicle pair
{
  "design": "two independent non-targeting hairpins, shLuc-1 and shLuc-2, both DMSO 72h, both 230927, all *_hg38_rmdup.tdf",
  "MEN1": [
    "GSM9032979",
    "GSM9032985"
  ],
  "MLL1": [
    "GSM9032981",
    "GSM9032987"
  ],
  "input": [
    "GSM9032977",
    "GSM9032983"
  ]
}
the five drug arm
{
  "design": "MV411 MEN1 and MLL1 ChIP at 230913, 50 nM 72h, matched inputs, all *_hg38_rmdup.tdf",
  "drugs": [
    "DMSO",
    "DS-1594b",
    "JNJ",
    "KO-539",
    "SNDX-5613"
  ],
  "why_it_matters": "this project has treated 'menin inhibitor' as one thing across roughly a dozen entries; five side by side in the index fusion is a direct test of that"
}
why the other two series failed
{
  "GSE260742": "within-condition replicates but mixed processing (filt.bam.tdf vs tdf)",
  "GSE127507": "uniform .deduped.tdf but no within-condition replicates at all"
}
mv411 identity verified
{
  "source": "Cellosaurus CVCL_0064, fetched 2026-09-23",
  "fusion": "KMT2A-AFF1 confirmed - 'Gene fusion; AFF1 + KMT2A', PubMed 1423625 and 8358709",
  "co_mutation": "FLT3 internal tandem duplication (ClinVar VCV000016270, PubMed 12529668)",
  "disease_annotation": "NCIt C9163 childhood acute MONOCYTIC leukemia; ORDO acute monoblastic/monocytic",
  "the_qualification": "right fusion, wrong lineage presentation - monocytic rather than the infant B-ALL this project is about. Closer to the subject than murine MLL-AF9 or MOLM-13, and not the disease itself."
}
download sizes measured
{
  "method": "HEAD request per supplementary file",
  "six_vehicle_pair_files_bytes": 1865230942,
  "total_gb": 1.87,
  "per_file_range_mb": [
    271,
    345
  ],
  "feasible": "one at a time, bin, delete - peak disk ~345 MB against C: 15.2 GB free",
  "GSM9032977": 344523744,
  "GSM9032979": 319514267,
  "GSM9032981": 323030012,
  "GSM9032983": 270906759,
  "GSM9032985": 307879181,
  "GSM9032987": 299376979
}
Still open
  1. ANSWERED: 1,865,230,942 bytes total for the six vehicle-pair files, 271-345 MB each. One at a time with bin-and-delete gives a ~345 MB peak, affordable on C:.
  2. Is the five-drug arm the better beat? DS-1594b, JNJ, KO-539, SNDX-5613 and MI-3454 in MV4-11 with matched inputs would test whether 'menin inhibitor' names one mechanism - an assumption this project has made in roughly a dozen entries and never examined. It may outrank the control that motivated finding this deposit.
  3. ANSWERED: Cellosaurus CVCL_0064 confirms the KMT2A-AFF1 fusion (PubMed 1423625, 8358709). It also records FLT3-ITD and annotates the line as childhood acute MONOCYTIC leukaemia - so it is the right fusion in the wrong lineage presentation, which any entry drawing on this deposit must say.
  4. Are shLuc-1 and shLuc-2 conservative enough to be the right control, or too conservative? They are independent integrations, so clonal drift is inside the comparison. If they show a gradient, distinguishing method artefact from clonal biology needs a third arm this deposit may not have.
Every source behind this entry (1)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.analysis

    GEO series GSE299100, titled 'UBR5 Mediates Menin Inhibitor Dependent Menin/KMT2A Complex Stability [ChIP-seq]', contains 64 samples in MV4-11 and its derivatives. Among them, MV411-shLuc-1 and MV411-shLuc-2 each provide a MEN1 ChIP, an MLL1 ChIP and an input, all DMSO 72h, all sequenced 230927 and all deposited as *_hg38_rmdup.tdf: GSM9032979/GSM9032985 (MEN1), GSM9032981/GSM9032987 (MLL1), GSM9032977/GSM9032983 (input). The series separately contains MV411 MEN1 and MLL1 ChIP at 230913 under DMSO, DS-1594b, JNJ, KO-539 and SNDX-5613 at 50 nM 72h with matched inputs, also as *_hg38_rmdup.tdf. MV4-11 carries the t(4;11) KMT2A::AFF1 fusion.

    What this evidence cannot show

    THIS IS A SURVEY OF TITLES AND FILENAMES, NOT A COMPUTATION. No file was downloaded, no bin was built, no genome build was verified beyond the hg38 string in the filenames, and the depositor's methods section has not been read. The uniform-processing claim rests on every relevant sample ending _hg38_rmdup.tdf - the same class of inference that failed in GSE260742, where the mixed suffixes were visible all along and nobody looked. shLuc-1 and shLuc-2 are independent hairpin integrations rather than technical replicates, so they carry clonal and off-target variation; that makes the control conservative but complicates a positive result. The MV4-11 karyotype is stated from the cell line's identity and was not verified against Cellosaurus in this cycle. Download sizes are unmeasured and D: has 9.7 GB free.

    Evidence that argues against this (3)
    • A vehicle pair built from two different hairpins may be too conservative to falsify anything. If shLuc-1 and shLuc-2 differ by clonal drift, a gradient between them could be real biology rather than a method artefact - and the control would then 'fail' for reasons that say nothing about regression to the mean. The clean test needs two aliquots of one culture, which this deposit does not obviously provide either, and I have presented a hairpin pair as though it settles the question when it is one step better than what was available rather than the thing itself.
    • Being in KMT2A::AFF1 is presented here as an unambiguous gain and it cuts both ways. Every prior chromatin result in this thread is in KMT2A::MLLT3, so a control run in a different fusion tests the METHOD in a setting where the underlying results were never measured. If the gradient turns out flat in MV4-11, that could mean the method is sound, or that MV4-11 simply behaves differently - and disentangling those needs the drug contrast run in MV4-11 too, which doubles the work this entry implies.
    • The five-drug comparison is the more valuable half of this deposit and it is buried in the second half of the summary. DS-1594b, JNJ, KO-539, SNDX-5613 and MI-3454 in one line with matched inputs would let this project test an assumption it has made in perhaps a dozen entries - that 'menin inhibitor' names one mechanism - and no beat has ever been dispatched on it. Framing the deposit primarily as a control for an internal methods question may be the wrong priority.
Connected to