start / graph / ana-q124-in-gse260742-the-filename-and-not-the-title-names-the-culture-and-there-are-four-murine-cultures

AnalysisstrongQ124-the-timecourse-and-the-pcgf1-knockout

GSE260742's murine H3K79me2 samples come from FOUR distinct cultures - F3, F3_NT, NT and the PCGF1-KO clone C1 - and the titles name only two of the distinctions; 'Timecourse Control' is the F3 line, and 'NTvKO Control' covers two different filename families that are not the same material

also called four murine cultures, not two · F3_NT and NT are different · Timecourse Control is F3

In this dataset the sample titles and the file names disagree about which batch of cells each experiment used. Reading the file names shows four separate cultures where the titles suggest two, and it settles a question the previous cycle had to leave open about whether one experiment used the same cells as another.

The same thing, for a scientist

**Q122's section 2a established that in GSE260742 the filename carries information the title does not**, and concluded that `NT` is the depositors' label for the non-targeting-sgRNA arm of the PCGF1-knockout experiment while `C1`/`CLONE1` is the knockout clone. **That is right and incomplete in two ways, both of which change what can be compared to what** [E1]. **First: `Timecourse Control` is the F3 line, and the characteristics block plus the filename settle it independently.** All six timecourse records carry `tissue: Blood`, `cell line: MLLAF9`, `cell type: Myeloid`, `Mus musculus` / taxid 10090 - the identical block as the `MLLAF9 Control` F3 samples Q122 measured - and their filenames are `ChIP_K79_**F3**_TIMECOURSE_*` and `ChIP_INPUT_**F3**_TIMECOURSE_..`. **So the timecourse is the F3 culture family in a separate experiment batch with its own vehicle (GSM8123452, 24 h) and its own input (GSM8123451), not a separate cell line.** This beat's dispatch brief anticipated that the cell system might be a bounded absence; it is not. **Second: `F3_NT` and bare `NT` are two different filename families, both appearing under `NTvKO Control` titles**, so there are **two distinct non-targeting-control H3K79me2 DMSO tracks** in this series and the titles do not distinguish them: | accession | `!Sample_title` | basename | mark | |---|---|---|---| | GSM8123386 | MLLAF9 Control - DMSO H3K79me2 **rep2** | `ChIP_K79_`**`F3_NT`**`_DMSO.tdf` | H3K79me2 | | **GSM8123438** | **NTvKO Control** - DMSO H3K79me2 | `ChIP_K79_`**`NT`**`_DMSO.tdf` | H3K79me2 | | GSM8123403 | NTvKO Control - DMSO H2AK119ub **paired** | `ChIP_K119UB_`**`F3_NT`**`_DMSO.filt.bam.tdf` | H2AK119ub | | GSM8123420 | NTvKO Control - DMSO H2AK119ub **rep2** | `ChIP_K119UB_`**`NT`**`_DMSO.tdf` | H2AK119ub | | GSM8123404 | NTvKO Control - DMSO input **paired** | `ChIP_INPUT_`**`F3_NT`**`_..filt.bam.tdf` | input | | GSM8123416 | NTvKO Control - DMSO **input3** | `ChIP_INPUT_`**`NT`**`_..tdf` | input | **The consequence is a gain, not a hazard.** Q122's criterion C3 measured a between-culture spread from **two** murine cultures. This beat adds a third (`NT`) and a fourth (`C1`), so the murine spread of `V` at the fusion's own sites is now measured over three non-knockout cultures: **F3 0.7734, F3_NT 0.8318, NT 0.8156, spread 0.0584** - and pre-registered criterion **P5 FIRES**, because 0.0584 is far below the 0.2089 cross-partner difference `hyp-q120` rests on. **`NT` lands between the two Q122 cultures**, so a third culture did not widen the spread at all. **No number from these comparisons is called a technical replicate, an SD or a standard error.** `F3_NT`, `NT` and `C1` all carry a transduction and a selection step, so an absolute difference among them is an **upper bound on technical noise and a lower bound on nothing.**

Go deeper

What this changes for anyone comparing numbers across these beats

culture	filename family	what has measured it	`V` at `S_FLAG` under VTP-50469 48 h
F3, the parental FLAG-MLLAF9 line	`_F3_`	Q117 (FLAG, KDM2B, BCOR, RING1B), Q122 (H3K79me2, H2AK119ub)	0.7734
F3_NT, F3 with a non-targeting guide	`_F3_NT_`	Q120 (MLL1), Q122 (H3K79me2)	0.8318
NT, the non-targeting arm of the knockout experiment	`_NT_`	this beat (H3K79me2)	0.8156
C1 / CLONE1, the PCGF1-knockout clone	`_C1_`, `_CLONE1_`	this beat (H3K79me2)	0.7990
F3_TIMECOURSE, F3 in the kinetic experiment	`_F3_TIMECOURSE_`	this beat (H3K79me2)	0.6790 at 24 h, and criterion K4 shows this batch is not on the F3 batch's scale

Q120's ARM 2 is the case worth flagging to its owner. Q122 already reported that Q120's FLAG tracks are FLAG_F3_* and its MLL1 tracks are MLL1_F3_NT_*, so its stated claim of 'the exact human reagent against a FLAG ground truth, in the same nuclei' is the parental line against its non-targeting-guide derivative. This beat adds that the derivative in question is F3_NT and not NT, which are themselves different, so the series contains three tiers of relatedness - same culture, sibling derivative, and a differently-named derivative - and the titles distinguish none of them. Reported, not edited: that entry is not this beat's.

The reusable form

When a deposit's titles and its filenames disagree about grouping, the filename is usually closer to the bench. A title is written for a reader at submission time and is edited for consistency; a supplementary filename is emitted by the pipeline and carries whatever the sample was actually called in the lab. In this series the titles were normalised into three conventions - MLLAF9 Control, NTvKO Control and Timecourse Control - and the filenames preserve five culture labels, one of which (F3_TIMECOURSE) has no title of its own at all.

So transcribe the filename beside the accession and the title, always, and treat a shared title as evidence of nothing about shared material. This beat's pre-registration transcribes all fourteen basenames for exactly that reason, and Q122's did the same for its fifteen.

Facts and figures

absence class	measured-not-curated
absence class note	The grouping information is PUBLISHED - it is in the !Sample_supplementary_file_1 field of every record - and is absent from the field any database or reader would join on, the title. So it is structurally invisible to anyone who reads titles, which is what Q120 did and what led it to describe two cultures as 'the same nuclei'. Closing it costs one grep over a file this project has fetched four times, and it stays measured-not-curated rather than measured-not-analysed because no analysis is required, only reading the right field.
absence class assigned	2026-09-23 (Q124)

murine culture families by filename token

{
  "F3": "the parental FLAG-MLLAF9 line; measured by Q117 and Q122",
  "F3_NT": "F3 carrying a non-targeting sgRNA; measured by Q120 (MLL1) and Q122 (H3K79me2)",
  "NT": "the non-targeting-sgRNA arm of the PCGF1-knockout experiment; measured by this beat",
  "C1 / CLONE1": "the PCGF1-knockout clone; measured by this beat",
  "F3_TIMECOURSE": "F3 in the kinetic experiment, with its own vehicle and input and no title of its own; measured by this beat, and criterion K4 shows it is not on the F3 batch's scale"
}

three culture spread of V at S FLAG

{
  "F3": 0.7734,
  "F3_NT": 0.8318,
  "NT": 0.8156,
  "max_minus_min": 0.0584,
  "threshold_from_q120": 0.2089,
  "criterion": "P5",
  "verdict": "FIRES"
}

How this could bite us back

A cross-culture spread is this project's only available noise estimate for these unreplicated measurements, and it is computed across labels whose physical meaning is inferred from filenames. If F3_NT and NT are the same culture, the three-culture spread of 0.0584 is really a two-culture spread and P1's threshold - which is Q122's F3-against-F3_NT difference - is measuring something narrower than claimed.
Conversely if the labels mean what they appear to, then Q120's claim to have measured two antibodies 'in the same nuclei' is a claim about a parental line and a derivative, and the masking numbers built on it inherit a between-culture component that nothing has bounded.

Still open

Are F3_NT and NT the same culture written two ways, or two cultures? Nothing in the 158 records settles it, and their H3K79me2 DMSO tracks could be compared directly - both are binned and on disk at C:/Temp/q122 and C:/Temp/q124 - which would at least say whether they are the same material to the resolution of a ChIP.
The H2AK119ub NTvKO arm has three replicate conventions using two filename families (F3_NT/F3_C1 for 'paired', bare NT/CLONE1 for rep1 and rep2). Is 'rep1' against 'rep2' a technical replicate pair, given that they share a filename family where 'paired' does not?
Does any other series this project uses have the same title-versus-filename divergence? GSE127507 was never checked for it, and Q118's and Q120's cross-line comparisons there rest on titles.

Every source behind this entry (1)

E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
In GSE260742, the six Timecourse samples (GSM8123447-452) carry the characteristics block tissue: Blood, cell line: MLLAF9, cell type: Myeloid, Mus musculus taxid 10090, and supplementary filenames of the form GSM####_ChIP_K79_F3_TIMECOURSE_*.filt.bam.tdf and GSM8123451_ChIP_INPUT_F3_TIMECOURSE_..filt.bam.tdf; and the title convention NTvKO Control covers two distinct filename families, F3_NT (GSM8123386 titled 'MLLAF9 Control - DMSO H3K79me2 rep2', GSM8123403 and GSM8123404 titled 'NTvKO Control - ... paired') and bare NT (GSM8123438, GSM8123416, GSM8123420), so the series contains two different non-targeting-control H3K79me2 DMSO tracks. The within-sample H3K79me2 share ratio at the FLAG-DMSO top-0.1% set under 500 nM VTP-50469 48 h is 0.7734 in F3, 0.8318 in F3_NT and 0.8156 in NT, a three-culture spread of 0.0584 against the 0.2089 cross-partner difference hyp-q120 rests on.
GEO series GSE260742 (Gilan Lab, Monash), per-sample SOFT records: 158 samples, 419,824 characters, fetched 2026-09-23 · NCBI Gene Expression Omnibus 2026
What this evidence cannot show
The mapping from a filename token to a physical culture is an INFERENCE from the depositors' naming, not a statement the records make: nothing in the 158 records says that F3_NT and NT are different material, and the alternative reading - that they are the same culture written two ways by two pipeline runs - is not excluded by anything here. The three-culture spread of 0.0584 is an absolute difference among cultures that each carry a transduction and a selection step, so it is an UPPER bound on technical noise and not an SD; n = 1 per culture. Two of the three V values are read from Q122's committed results.json rather than re-derived. The F3_TIMECOURSE batch is EXCLUDED from the three-culture spread because criterion K4 shows it is not on the same scale as the F3 batch.
Evidence that argues against this (1)
- NCBI Gene Expression Omnibus 2026

Connected to

supports →In murine MLL-AF9, knocking out PCGF1 changes the menin inhibitor's effect on H3K79me2 by 0.0166 of a share ratio and the DOT1L inhibitor's by 0.0101 - both smaller than the 0.0584 two control cultures of the same line already differ by, so the mark's drug response runs outside PRC1.1
identifies the NT and C1 cultures the PCGF1 arm compares and supplies the third-culture value (NT, 0.8156) that makes the three-culture spread 0.0584
refutes →Hypothesis: H3K79me2 at a KMT2A fusion's own sites survives menin inhibition when the C-terminal partner contacts DOT1L directly (MLLT3/AF9) and decays when it does not (AFF1) - so the mark's response is a property of the partner and not of menin, which predicts the opposite of the direct-contact intuition and is testable in one line
a third murine culture lands between the two Q122 measured (F3 0.7734, NT 0.8156, F3_NT 0.8318), so the murine spread stays at 0.0584 against the 0.2089 cross-partner difference - the hypothesis's effect survives a harder test than two cultures gave it, and the test does not widen
supports →The only deposited H3K79me2 kinetic series for a menin inhibitor is non-monotone - a minimum at 16 h and a recovery at 24 h that survives common support and two rank statistics - and its 24 h point shows MORE redistribution than the 48 h main-arm sample, so the 48 h/72 h asymmetry in cycle 82's F stands and F keeps the words lower bound
establishes that Timecourse Control is the F3 line in a separate batch, which is what makes the batch-scale test K4 the right question and its failure interpretable
← supportsSix GSE260742 samples titled 'NTvKO ... H2AK119ub rep1' carry ChIP_FLAG filenames AND a ChIP_FLAG string in the TDF header their own depositors' igvtools wrote - two of three independent sources say FLAG, the shape test is near-indecisive, and the channel of half the replicate arm cannot be settled from the deposit
the same title/filename disagreement, now about the ANTIBODY rather than the culture