Cell type or the fusion? MV4-11 LOOSENS the confound the graph's own lineage split was built on - but its founding paper calls it biphenotypic
also called Q60 DepMap cell-type contrast · same fusion, two cell types
To ask whether a gene matters because of the leukaemia's cell type or because of its fusion gene, you need the same fusion growing in two different kinds of cell. One cell line, MV4-11, was meant to be that, and it was missing from this project's earlier comparison. Adding it suggests menin and DOT1L are needed in both cell types equally - but the paper that first described MV4-11 calls it a mixed B-and-myeloid line, so the two cell types are not as separate as the design needs.
The same thing, for a scientist
The repo's existing lineage-confound analysis splits the KMT2A-rearranged DepMap arm into a myeloid half and a B-lineage half. In that panel the myeloid lines all carry KMT2A::MLLT3 or KMT2A::AFDN and the B-lineage lines all carry KMT2A::AFF1, so lineage and partner move together and a 'lineage effect' is not separable from a partner effect. MV4-11 (Cellosaurus CVCL_0064, DepMap ACH-000045) carries KMT2A::AFF1 and was absent from that panel. **LINEAGE CORRECTED 2026-09-21 (Q68): Cellosaurus annotates MV4-11 as childhood acute monocytic leukaemia, but the paper that first characterised the line's growth-factor requirements calls it, in its own words, 'the biphenotypic B-myelomonocytic leukemia (MV4-11) cells' (PMID 3500218), distinguishing it in the same sentence from AML-193, 'the monocytic leukemia' line. The myeloid arm of this contrast is therefore a B-myelomonocytic line, and the separation of cell type from fusion partner is looser than 'breaks the confound' implies.** Re-running the contrast with MV4-11 added, over the sixteen highest-degree core genes plus MEN1, KMT2A, RBBP5 and BRD4, gives three nested contrasts and one dominant result: six of twenty genes are floor-compressed (depleted in >=90% of all 1,258 screened lines), and for those genes no cell-type question can be asked of DepMap at all.
The three contrasts and what each can and cannot say
A. KMT2A-rearranged (any partner) minus all 1,258 lines. Is this a leukaemia-context dependency at all? MEN1 -0.901, FLT3 -0.817, KMT2A -0.605, PSIP1 -0.582, DOT1L -0.457. Nothing else in the panel exceeds -0.33. 1
B. KMT2A-r myeloid minus KMT2A-r B-lineage. This is the repo's existing split and it remains confounded by partner. FLT3 +0.283 and MLLT1 +0.260 are the largest (B-lineage more dependent); MEN1 is +0.008, i.e. flat. 1
C. KMT2A::AFF1 myeloid (MV4-11) minus KMT2A::AFF1 B-lineage (SEM, SEMK2, RS4;11). Same fusion, different cell type. This is n = 1 against n = 3 and is a look, not a test. MEN1 -0.168 and DOT1L -0.143 are near zero, so the two central dependencies do not move with cell type when the fusion is held fixed. 1
Why contrast C matters more than its n. Every other comparison in this graph that claims to separate cell type from fusion has to argue the point; this one holds the fusion constant by construction.
Floor compression comes first, and it removes six genes from the question
WDR5 (98.3% of all lines depleted below -0.5), KAT8 (99.5%), RBBP5 (97.7%), BRD4 (94.8%), ELL (93.9%) and PAF1 (93.4%) are pan-essential across the screened cancer cell lines. For these six, a small index-minus-comparator difference is uninterpretable rather than small, because there is no dynamic range left in which cell-type selectivity could appear. 1
This is not a marginal caveat for this beat: six of the sixteen worklist genes are in this class. The honest answer to 'which cell type does WDR5 matter in' from DepMap is 'every one that has been screened', which is a statement about cell viability in culture and carries no cell-type information whatever. Any cell-type claim about those six has to come from conditional knockouts in defined haematopoietic compartments, not from a cancer dependency screen.
The panel, and what was dropped
Eleven KMT2A-rearranged lines were matched, each verified against Cellosaurus for fusion partner and disease: MV4-11 (KMT2A::AFF1, monocytic), THP-1, MOLM-13, MOLM-14, NOMO-1 (KMT2A::MLLT3, monocytic/AML), SHI-1, OCI-AML-2 (KMT2A::AFDN), SEM, SEMK2, RS4;11 (KMT2A::AFF1, B-ALL), HB11;19 (KMT2A::MLLT1, B-ALL).
Named and dropped, per brief rule 78: KOPN-8 (CVCL_1866, KMT2A::MLLT1 B-ALL) and ML-2 (KMT2A::AFDN AML) are absent from this DepMap release; Mono-Mac-1 (CVCL_1425) is present and monocytic but carries no gene-fusion annotation in Cellosaurus, so it was left out rather than assigned a partner. The dropped set is not concentrated in one arm, but it does remove one of the two non-AFF1 B-lineage lines.
What this does NOT establish
A DepMap gene effect is a statement about an immortalised line in culture over about three weeks. It is not a statement about the normal cell type the line resembles, and it is not a statement about a patient's blasts. It also measures loss of the whole gene: for AFF1, MLLT1, MLLT3 and MLLT10 the library's guides cut the wild-type allele and the fusion allele together, so a null result is not evidence that the fusion's 3' moiety is dispensable, and the striking AFF1 null (0.2% of all lines depleted; KMT2A-r mean -0.257) has to be read that way. Finally, MV4-11 carries FLT3-ITD, which is a second difference from the B-lineage KMT2A::AFF1 lines and makes contrast C uninterpretable for FLT3 itself.
Correction, 2026-09-21 (Q68): the line is biphenotypic, so this loosens the confound rather than breaking it
What was withdrawn. The title verb 'breaks', and the description of MV4-11 as 'monocytic AML' without qualification.
Why. PMID 3500218 (Santoli et al., J Immunol 1987) is the paper that established MV4-11's GM-CSF dependence, and in its own text it calls MV4-11 'the biphenotypic B-myelomonocytic leukemia' line while calling AML-193 in the same sentence 'the monocytic leukemia' line. Cellosaurus's NCIt annotation (C9163, childhood acute monocytic leukemia) is curated and, as this entry's own caveat says, coarse.
What this does and does not change. It does NOT restore the original confound: MV4-11 still carries KMT2A::AFF1 in a non-B-lineage-restricted background, which the previous panel had no example of, and the floor-compression result (six of twenty genes uninterpretable) is untouched. It DOES mean contrast C is a comparison between a B-myelomonocytic line and three B-lineage lines, so the cell-type axis is partly within one mixed phenotype and the invariance of MEN1 (-0.168) and DOT1L (-0.143) across it is weaker evidence for cell-type invariance than it looks. Combined with n = 1 against n = 3 and the FLT3-ITD, contrast C should be read as consistent-with rather than demonstrating.
And it strengthens the entry's own recommendation: exp-same-fusion-two-cell-types-allele-selective-degron argues for an isogenic knock-in pair, and the biphenotypic annotation is a better argument for that than the FLT3-ITD is.
| script | analyses/q60-cell-type-versus-fusion/run.py |
|---|---|
| status | run |
| result summary | Six of twenty genes floor-compressed and therefore uninformative about cell type. MEN1 is the only large KMT2A-r-selective dependency and it is cell-type-flat. FLT3 and MLLT1 are the two genes whose dependency is larger in the B-lineage half. |
| n lines screened | 1258 |
| kmt2ar lines matched | 11 |
| kmt2ar lines named and dropped | 3 |
inputs
[ "Open Targets GraphQL depMapEssentiality", "Cellosaurus cell-line records", "HGNC REST symbol-to-Ensembl dictionary resolved 2026-09-20" ]
floor compressed genes
[ "WDR5", "KAT8", "RBBP5", "BRD4", "ELL", "PAF1" ]
contrast A kmt2ar minus all lines — the numbers behind the chart
{
"MEN1": -0.901,
"FLT3": -0.817,
"KMT2A": -0.605,
"PSIP1": -0.582,
"DOT1L": -0.457,
"AFF1": -0.325,
"MLLT1": -0.311,
"ELL": -0.183,
"MLLT10": -0.156,
"DPY30": -0.12,
"BRD4": -0.117,
"KMT2B": -0.105,
"IGF2BP3": -0.087,
"RBBP5": -0.053,
"KAT8": -0.036,
"PAF1": 0.012,
"MLLT3": 0.054,
"AFF4": 0.08,
"WDR5": 0.081,
"ASH2L": 0.157
}contrast C same fusion myeloid minus blineage — the numbers behind the chart
{
"KMT2A": -0.757,
"FLT3": -0.525,
"MLLT1": -0.316,
"PSIP1": -0.201,
"DPY30": -0.179,
"MEN1": -0.168,
"DOT1L": -0.143,
"WDR5": -0.13,
"KAT8": -0.124,
"AFF1": -0.093,
"MLLT10": -0.089,
"MLLT3": -0.004,
"AFF4": -0.008,
"RBBP5": 0.066,
"IGF2BP3": 0.126,
"ASH2L": 0.147,
"KMT2B": 0.168,
"BRD4": 0.285,
"PAF1": 0.418
}- A target that is pan-essential in 99% of screened lines (KAT8) has no therapeutic index argument available from this data; its apparent KMT2A-r dependency is the floor, not selectivity.
- If the only evidence a gene 'matters in myeloid KMT2A-r disease' is the old lineage split, that evidence is partner-confounded and the drug rationale inherits the confound.
- MEN1's cell-type flatness cuts both ways: it supports menin inhibition across lineages, and it removes any argument that normal B-lineage or myeloid progenitors would be spared.
- Does a second KMT2A::AFF1 line with an unambiguous, single-lineage myeloid phenotype exist anywhere that could be added to DepMap? MV4-11's founding paper calls it biphenotypic B-myelomonocytic (PMID 3500218), so contrast C is currently n = 1 AND mixed-phenotype.
- For the six floor-compressed genes, is there any conditional knockout in a defined haematopoietic compartment that could supply the cell-type answer a dependency screen cannot?
- Would an allele-selective degron on the fusion protein, rather than whole-gene CRISPR, change the AFF1, MLLT1, MLLT3 and MLLT10 nulls?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
In DepMap CRISPR gene effect fetched live on 2026-09-20 across 1,258 screened lines, six of twenty core KMT2A-axis genes (WDR5, KAT8, RBBP5, BRD4, ELL, PAF1) are depleted in at least 93% of all lines, and MEN1 shows the largest KMT2A-rearranged-versus-all-lines difference (-0.901) while moving only -0.168 between KMT2A::AFF1 myeloid and KMT2A::AFF1 B-lineage cells.
Open Targets Platform - DepMap CRISPR gene essentiality (depMapEssentiality), queried by Ensembl gene id · Open Targets Platform GraphQL API 2026What this evidence cannot show
A cancer cell line dependency screen. Eleven KMT2A-rearranged lines, seven myeloid and four B-lineage, so one line moves a mean; contrast C is n = 1 against n = 3. Lineage is the Cellosaurus disease annotation, which is coarse. Gene effect is whole-gene loss, so fusion and wild-type alleles are cut together for the 3' partner genes.
Evidence that argues against this (1)
- The repo's own prior analysis (analyses/lineage-confound-myeloid-tf-dependencies/) expressed the same contrast in standard deviations of the comparator arm rather than in raw gene effect and reported PSIP1's myeloid-minus-B-lineage gap as -1.06 where this analysis in raw units gives -0.231. Both are correct in their own frame; the brief's cycle-45 rule applies, that a z-score divided by a narrow comparator SD is close to a ranking of narrowness rather than of effect size. Report both or say which you mean. oncofusion-knowledge-graph 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
MV4-11 is a KMT2A::AFF1 cell line of MONOCYTIC lineage - karyotype 48,XY,t(4;11)(q21;q23), gene fusion KMT2A-AFF1, NCIt C9163 childhood acute monocytic leukaemia - and it also carries an FLT3 internal tandem duplication.
Cellosaurus cell line MV4-11 (CVCL_0064) · Cellosaurus, SIB Swiss Institute of Bioinformatics 2026What this evidence cannot show
Cellosaurus disease annotation is curated from the depositor and the literature and is coarse. The FLT3-ITD is annotated from PMID 12529668. Read back live on 2026-09-20.
Evidence that argues against this (1)
- PUBLISHED CONFLICTING RESULT about the lineage of the line the whole contrast turns on. Cellosaurus records MV4-11 as childhood acute monocytic leukaemia, but the paper that first characterised the line's growth-factor requirements calls it, in its own words, 'the BIPHENOTYPIC B-MYELOMONOCYTIC leukemia (MV4-11) cells', alongside AML-193 which it separately calls monocytic. If MV4-11 is biphenotypic, then the contrast this analysis was built to create - one myeloid KMT2A::AFF1 line against three B-lineage KMT2A::AFF1 lines - has a B-myelomonocytic line on the myeloid side and is a weaker separation of cell type from partner than it reads. That does not restore the original confound, but it does mean the entry's claim to BREAK the confound overstates what one biphenotypic line can do. Santoli D et al. 1987
the analysis that motivates the three evidence classes: six of twenty core genes are floor-compressed and carry no cell-type information
- related to →KMT2A / MLL1 — the gene that breaks
the index fusion's 5' gene; KMT2A-r arm mean gene effect -0.811 against -0.206 across all 1,258 lines
MEN1 has the largest KMT2A-rearranged selectivity in the panel (-0.901) and moves only -0.168 when the fusion is held constant and cell type varies
MV4-11 (myeloid) against SEM, SEMK2 and RS4;11 (B-lineage) holds this fusion constant while varying cell type
MV4-11 is annotated as childhood acute MONOCYTIC leukaemia and is the myeloid arm of the same-fusion contrast
SEM, SEMK2 and RS4;11 are B-lineage KMT2A::AFF1 lines and are the B-lineage arm of the same-fusion contrast
the worked example of floor compression: 99.5% of 1,258 lines depleted
floor-compressed at 93.4% of lines, so DepMap carries no cell-type information for PAF1
floor-compressed at 98.3% of lines; the KMT2A-r arm is marginally LESS dependent than average
second largest KMT2A-r selectivity in the panel, and the one gene for which contrast C is uninterpretable because MV4-11 is FLT3-ITD
arises from the same-fusion, two-cell-type contrast
floor-compressed at 93.9% of lines, so neither the literature nor the screen carries cell-type information
AFF4 is dispensable in 99.6% of screened lines and in every KMT2A-rearranged line
the only positive, non-floor-compressed contrast in the worklist: KMT2A-r needs ASH2L LESS than the average line (+0.157)
LINEAGE CORRECTED: MV4-11's founding characterisation calls it 'the biphenotypic B-myelomonocytic leukemia' line (PMID 3500218), so contrast C is a mixed-phenotype line against three B-lineage lines - which is a better argument for the isogenic knock-in pair than the FLT3-ITD is