Menin dependence varies 2.83x across KMT2A::AFF1 lines and the spread survives at three donors - but RS4;11 is last on KMT2A itself, so the heterogeneity is largely fusion-dependence heterogeneity, and DAXX is the one gene decoupled from it
also called q177 · does the menin spread survive · RS4;11 is the KMT2A outlier too
Last analysis found that two cell lines with the same fusion needed menin very differently, which would be awkward for the idea that the fusion works through menin. Fetching every available line with that fusion, the difference holds. But the odd line out also barely needs the fusion gene itself - it is the least fusion-dependent line on the list - so it needs less menin because it needs less of the whole machine, not because menin works differently there. That largely rescues the idea. DAXX behaves differently: it is needed evenly, regardless of how much a line depends on the fusion.
The same thing, for a scientist
q176 recorded a 2.84× spread in menin dependence between two KMT2A::AFF1 donors as *"the sharpest quantitative challenge to the menin-interface hypothesis currently in the corpus"*, at n=2. This fetches every KMT2A::AFF1 line DepMap has and then applies the control that decides it. 🟠 **The spread survives, and it is not lineage.** Four of the five AFF1 lines have a CRISPR screen (**ALL-PO has none — missing, not weak**) [E1]. Over three donors MEN1 spreads **2.83×**, essentially unchanged from the two-donor 2.84×, against DAXX's **1.18×**. The pre-registered lineage explanation fails: MV4-11, the myeloid/biphenotypic line, sits at **−1.1980**, inside the B-lineage range of −1.667 to −0.463 [E1]. 🟢 **But RS4;11 is the KMT2A outlier too, and that largely rescues the hypothesis.** The control fetched alongside is the dependency on **KMT2A itself** [E2]: | line | KMT2A | MEN1 | DOT1L | DAXX | |---|---:|---:|---:|---:| | MV4-11 | **−1.2890** | −1.1980 | −0.6534 | −0.9674 | | SEMK2 | −0.8771 | **−1.6671** | −0.6154 | −0.7979 | | SEM | −0.5923 | −0.9593 | −0.5480 | −0.9475 | | **RS4;11** | **−0.1282** | **−0.4632** | −0.3675 | −0.8214 | **RS4;11 ranks 4 of 4 on KMT2A, 4 of 4 on MEN1 and 4 of 4 on DOT1L.** It is not a line with an unusual menin relationship; it is a line that depends less on the fusion axis as a whole. **Menin dependence tracks fusion dependence (Spearman +0.80), which is what a required cofactor should do** — so q176's challenge is substantially weakened and should not be cited as standing. 🔵 **The positive control works and puts menin second.** **DOT1L tracks KMT2A at Spearman +1.00** (Pearson +0.9554) — perfect rank agreement — confirming the method can detect a fusion-axis relationship on these four points, and showing menin's +0.80 is *slightly looser* than the canonical elongation cofactor's [E2]. 🟠 **DAXX is the one gene decoupled from the fusion axis (Spearman +0.40), and that is now its most distinctive property.** It is required at −0.80 to −0.97 in every AFF1 line regardless of how much that line needs KMT2A. For a chaperone that makes sense; it also means **DAXX dependence is not evidence of a fusion-specific mechanism** — an evenly-required gene is what a general requirement looks like. ATRX is *anti*-correlated (−0.80), which on n=4 is most likely noise and is recorded rather than interpreted. 🔴 **The normalisation does NOT flatten, and I am not claiming it does.** MEN1/KMT2A spreads **3.89×**, slightly *more* than the raw 3.60×. But RS4;11's denominator is **−0.1282**, near zero, so the ratio is least trustworthy exactly where it matters. **The ordering is the evidence here; the ratio is not.**
What would have gone wrong without the KMT2A control row
The MEN1 numbers alone say: two lines with the same fusion differ 2.83-fold in menin dependence, over three donors, not explained by lineage. Read on their own that is a striking result and a publishable-sounding challenge to a central hypothesis.
The KMT2A row turns it into something much weaker, and it cost one extra gene in the same query. RS4;11 is last on KMT2A, MEN1 and DOT1L simultaneously. A line that is only weakly fusion-dependent is expected to be weakly dependent on everything the fusion needs, and menin dependence tracking fusion dependence is a prediction of the interface model, not a problem for it.
The general rule this supports: when a subject shows unexpected variation across samples, fetch the variation in the thing it is supposed to depend on, in the same query. Here the control was one extra Ensembl id. This corpus has a standing habit of asking whether an effect is confounded; the cheaper and often more decisive question is whether the effect is simply inherited from a quantity nobody measured.
What survives as a challenge, and what n=4 cannot settle
Weakened but not dead. Spearman +0.80 on four points is one discordant pair away from +0.40. SEMK2 is the discordance: it is most menin-dependent (−1.6671) while only second on KMT2A (−0.8771), and MV4-11 is most KMT2A-dependent (−1.2890) while second on MEN1. So menin dependence is not a clean function of fusion dependence even in these four lines — the two most fusion-dependent lines swap places on menin.
What n=4 cannot settle. Three donors, one of which (SEM/SEMK2) supplies two lines that disagree with each other by 0.71 Chronos units on MEN1 — a within-donor spread nearly as large as the between-donor spread this analysis is about. That single fact bounds how much any of this can mean: if the same donor's two sublines differ by 0.71, a 0.85 difference between donors is not obviously biology. No entry should quote the 2.83× without it.
ALL-PO remains unscreened. This corpus already holds an entry noting ALL-PO is in DepMap and has never been screened; that is confirmed here. It is an infant KMT2A::AFF1 line, i.e. the closest available model to the index population, and its absence is the single biggest gap in this comparison.
| script | analyses/q177-menin-dependence-across-every-aff1-line/{run,vs_kmt2a}.py |
|---|---|
| status | executed |
| answers open question from | ana-q176-the-index-arm-contains-a-non-index-fusion-and-menin-fails-on-variance-not-on-effect |
| depmap via | Open Targets GraphQL depMapEssentiality, fetched 2026-09-24 |
| men1 donor spread fold | 2.835 |
| daxx donor spread fold | 1.178 |
| within donor spread men1 sem vs semk2 | 0.7078 |
| verdict | the spread survives but is largely inherited from fusion dependence; q176's challenge to the menin-interface hypothesis is substantially weakened and should not be cited as standing |
aff1 lines scored
[ "SEM", "SEMK2", "RS4;11", "MV4-11" ]
aff1 line unscreened
[ "ALL-PO" ]
men1 per line — the numbers behind the chart
{
"SEM": -0.9593,
"SEMK2": -1.6671,
"RS4;11": -0.4632,
"MV4-11": -1.198
}kmt2a per line — the numbers behind the chart
{
"SEM": -0.5923,
"SEMK2": -0.8771,
"RS4;11": -0.1282,
"MV4-11": -1.289
}daxx per line — the numbers behind the chart
{
"SEM": -0.9475,
"SEMK2": -0.7979,
"RS4;11": -0.8214,
"MV4-11": -0.9674
}spearman vs kmt2a — the numbers behind the chart
{
"DOT1L": 1,
"MEN1": 0.8,
"DAXX": 0.4,
"ATRX": -0.8
}rs411 rank of 4 — the numbers behind the chart
{
"KMT2A": 4,
"MEN1": 4,
"DOT1L": 4,
"DAXX": 3,
"ATRX": 2
}measurements — the numbers behind the chart
[
{
"quantity": "fold",
"subject": "MEN1 dependency spread across KMT2A::AFF1 donors",
"universe": "3 donors, 4 scored DepMap AFF1 lines",
"value": 2.835
},
{
"quantity": "fold",
"subject": "DAXX dependency spread across KMT2A::AFF1 donors",
"universe": "3 donors, 4 scored DepMap AFF1 lines",
"value": 1.178
},
{
"quantity": "spearman",
"subject": "MEN1 dependency vs KMT2A dependency",
"universe": "4 KMT2A::AFF1 lines",
"value": 0.8
},
{
"quantity": "spearman",
"subject": "DOT1L dependency vs KMT2A dependency",
"universe": "4 KMT2A::AFF1 lines",
"value": 1
},
{
"quantity": "spearman",
"subject": "DAXX dependency vs KMT2A dependency",
"universe": "4 KMT2A::AFF1 lines",
"value": 0.4
},
{
"quantity": "chronos_delta",
"subject": "within-donor MEN1 difference, SEM vs SEMK2",
"universe": "one donor, two sublines",
"value": 0.7078
}
]- Get ALL-PO screened, or find any published CRISPR/shRNA menin result in it. It is a B-lineage KMT2A::AFF1 line and the closest available model to the infant index population, and it is the single biggest hole in this comparison.
- Bound the technical variance properly: SEM and SEMK2 are one donor and differ by 0.71 Chronos units on MEN1. Compute the within-donor difference for many genes to get a null for what a 0.85 between-donor difference is worth. Until that exists, the 2.83x should not be quoted without the 0.71 beside it.
- Does DAXX's decoupling from fusion dependence hold across the KMT2A-rearranged lines of other partners? If DAXX is evenly required everywhere it is a general requirement and not a fusion-specific one, which would deflate the q172 beat considerably.
- Repeat the KMT2A-control trick on this corpus's other dependency claims. Any gene whose cross-line variation was read as biology should be re-scored against the variation in KMT2A itself; the control costs one Ensembl id in the same query.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
Of the five DepMap lines carrying KMT2A-AFF1 (SEM, SEMK2, RS4;11, MV4-11, ALL-PO), four have a CRISPR gene-effect value and ALL-PO has none. Chronos MEN1 values are SEM -0.9593, SEMK2 -1.6671, RS4;11 -0.4632, MV4-11 -1.1980; collapsing SEM/SEMK2 to one donor gives three donors spreading 2.83-fold in magnitude (SEM -1.3132, MV4-11 -1.1980, RS4;11 -0.4632), against DAXX's 1.18-fold (-0.8727, -0.9674, -0.8214). The myeloid/biphenotypic line MV4-11 lies inside the B-lineage range, so the spread does not track lineage.
What this evidence cannot show
Four lines, three donors. Chronos scores are normalised per screen and carry technical variance that is not bounded here. Line-to-fusion assignment is taken from this repo's 2026-09-21 identity resolution rather than re-queried from Cellosaurus. 'Lineage' is a two-category split with one line on one side.
Evidence that argues against this (2)
- The two sublines of a SINGLE donor, SEM and SEMK2, differ by 0.71 Chronos units on MEN1 (-0.9593 vs -1.6671) - nearly as much as the 0.85 between-donor spread the analysis is about. That within-donor variation is the natural null for the between-donor claim, and it very nearly abolishes it.
- ALL-PO having no screen is not neutral: it is a B-lineage KMT2A::AFF1 line and arguably the closest model to the infant index population. Its absence means the four scored lines are a convenience sample, and the spread is estimated on whichever lines a screening programme happened to include.
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Fetched in the same query, Chronos KMT2A gene effect across the four scored AFF1 lines is MV4-11 -1.2890, SEMK2 -0.8771, SEM -0.5923, RS4;11 -0.1282. RS4;11 ranks 4 of 4 (least dependent) on KMT2A, MEN1 and DOT1L. Across the four lines, dependency correlates with KMT2A dependency at Spearman +0.80 (Pearson +0.7356) for MEN1, +1.00 (Pearson +0.9554) for DOT1L, +0.40 (Pearson +0.4977) for DAXX and -0.80 (Pearson -0.6353) for ATRX. The MEN1/KMT2A ratio spreads 3.89-fold, more than the raw MEN1 spread of 3.60-fold.
What this evidence cannot show
A correlation on four points. Spearman +0.80 on n=4 is a single swapped pair away from +0.40, and the analysis names the discordance: SEMK2 is first on MEN1 but second on KMT2A. The MEN1/KMT2A ratio is unstable because RS4;11's denominator (-0.1282) is near zero, which is precisely the line the argument turns on - the ordering is reported as the evidence and the ratio explicitly is not.
Evidence that argues against this (3)
- The ratio result runs AGAINST the entry's conclusion and is reported rather than dropped: normalising menin by fusion dependence does not flatten the spread, it widens it to 3.89x. If the near-zero denominator were not a known artefact this would be evidence that menin heterogeneity is NOT inherited from fusion dependence, i.e. the opposite conclusion.
- DOT1L correlating with KMT2A at Spearman +1.00 across four points has a one-in-twelve chance of arising from a random ordering, so the 'positive control works' reading is itself weakly supported. And a perfect correlation is what one would also see if all four values were driven by a shared per-line technical factor such as screen quality or proliferation rate, which would inflate every correlation in this table including MEN1's.
- ATRX's negative correlation (-0.80) is dismissed as noise because n=4, but the same n applies to MEN1's +0.80, which is not dismissed. Treating a coefficient as signal when it supports the conclusion and as noise when it does not is exactly the asymmetry this corpus has been caught in before; the honest position is that no coefficient here is individually reliable and only the RANK-4-on-everything observation about RS4;11 is robust.
answers q176's first open question: the 2.835x spread survives but RS4;11 is last on KMT2A too, so the challenge is largely inherited from fusion dependence
menin dependence tracks KMT2A dependence at Spearman +0.80 across four AFF1 lines, as a required cofactor should
DAXX is the one gene decoupled from fusion dependence (Spearman +0.40) - evenly required, which is what a general requirement looks like
DOT1L tracks KMT2A at Spearman +1.00 - the positive control that shows the method can see a fusion-axis relationship at n=4
answers q177's within-donor caveat: MEN1 is the 99.4th percentile of subline variation, so its dependency is unstable rather than patient-heterogeneous