Necessary is not explanatory: four different questions that a knockout answers as if they were one
also called requirement versus ordering · necessary but not sufficient · requirement is not specificity
If removing a protein stops a cancer, that proves the cancer needs it - and almost nothing else. It does not say the protein causes the cancer, that it acts first, that it is the slow step, or that it is what makes this cancer different from any other.
The same thing, for a scientist
Most mechanistic claims in this field rest on a **requirement experiment**: delete, knock down or degrade X, and the phenotype fails. That design answers exactly one question - *is X needed?* - and is routinely read as answering four. The KMT2A-fusion causal chain is built almost entirely out of requirement experiments, and every one of its documented breaks is an instance of one of these four confusions. This entry names them so an entry can state which it means.
The four questions and the experiment each one needs
| question | what it asks | the design that answers it | worked example in this disease |
|---|---|---|---|
| Requirement | does the phenotype fail without X? | knockout, knockdown, degron, inhibitor | menin is required for KMT2A-fusion leukaemia 1 |
| Ordering | does X change before Y? | acute removal plus a timecourse, ideally degron | the only one in the whole chain: elongation changes at 15 min, chromatin later 2 |
| Determinant | is X what makes this system different from a comparator that does not transform? | a paralogue or partner swap that holds X constant and changes the phenotype | KMT2B carries an equivalent menin motif and cannot make a transforming fusion; the CXXC/post-CXXC address is the determinant 3 |
| Rate-limiting | does the phenotype scale with the amount of X over the range a drug can reach? | allelic or pharmacological dose titration | not available for HOXA9 or MEIS1; germline Hoxa9 loss does not change MLL-AF9 incidence 4 |
A requirement experiment is cheap and the other three are not, which is why the literature is 90% column one and why a synthesis built from it silently claims columns two to four.
Why the confusion is systematic rather than sloppy
Three structural reasons, all visible in this disease:
- The word 'essential' does duty for all four. A DepMap common-essential gene is a requirement claim with no ordering, determinant or dose content at all, and this project has had to write three separate corrections about reading selectivity off it.
- A drug reads as a determinant because it works. Menin inhibition produces complete remissions 5. The inference therefore the menin contact is what makes the fusion oncogenic is the determinant claim, and it is refuted by a paralogue swap 3 - a result published in 2009 and rarely cited alongside the drug.
- Partial knockdown is silently treated as a dose titration. An shRNA that removes 70% of a protein and halves a phenotype licenses no EC50 and no statement about fractional occupancy, which is the quantity a medicinal chemist actually needs 4.
The asymmetry to remember: a requirement result is robust and narrow. It rarely turns out to be wrong; it routinely turns out to be less than what was claimed.
How to write it in an entry
- State the question class in the claim: 'required for' not 'drives', unless an ordering experiment exists.
- If the entry asserts an arrow between two nodes, say whether the arrow is measured or inferred, and put the system in the same sentence.
- A determinant claim needs a comparator that fails; name it. If no comparator exists, the claim is a requirement claim wearing a determinant's clothes.
- A rate-limiting claim needs a dose axis.
mechanism-rate-limiting-steps-and-compensationuses the word for six steps of this chain; only two of them have anything resembling a dose measurement. - The reverse error exists too. An absence of requirement is not an absence of role: Hoxa9-null mice still get MLL-AF9 leukaemia 4, and HOXA9 is still required in established human disease 6. Two experiments, two species, two disease stages, opposite headlines, both correct.
| applied to | mechanism-kmt2a-fusion-causal-chain |
|---|---|
| writing rule | state the question class in the claim; an arrow between nodes must say measured or inferred and name the system |
question classes
[
{
"class": "requirement",
"asks": "does the phenotype fail without X",
"design": "knockout, knockdown, degron, inhibitor",
"example_evidence": "E1"
},
{
"class": "ordering",
"asks": "does X change before Y",
"design": "acute removal plus timecourse",
"example_evidence": "E2"
},
{
"class": "determinant",
"asks": "is X what distinguishes this system from a non-transforming comparator",
"design": "paralogue or partner swap",
"example_evidence": "E3"
},
{
"class": "rate-limiting",
"asks": "does the phenotype scale with the amount of X over a druggable range",
"design": "allelic or pharmacological dose titration",
"example_evidence": "E4"
}
]chain arrow counts — the numbers behind the chart
{
"measured_orderings": 1,
"inferred_arrows": 4,
"refuted_as_serial": 2
}- A target chosen on requirement alone can be fully engaged and still not be the slow step, which is the commonest way a mechanistically correct drug underperforms.
- Reading a requirement as a determinant produces combination strategies that pair two agents acting on the same necessary-but-not-explanatory node, and expects synergy that the biology never promised.
- How many claims in this graph use a causal verb ('drives', 'causes', 'leads to') while citing only a requirement experiment? A verb audit against evidence design class is mechanically possible and has never been run.
- Is there a cheap surrogate for an ordering experiment when a degron is not available - for instance a washout of a reversible inhibitor with a dense early timecourse?
- Should the schema carry an explicit design-class field on each evidence item, so that a synthesis can compute what it is entitled to claim?
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Menin is required to initiate KMT2A-mediated leukaemogenesis and for maintenance of KMT2A-associated but not other oncogene-induced myeloid transformation, and acute menin ablation reverses the differentiation arrest.
The menin tumor suppressor protein is an essential oncogenic cofactor for MLL-associated leukemogenesis. · Yokoyama A, Somervaille TC, Smith KS, Rozenblatt-Rosen O, et al. · Cell 2005What this evidence cannot show
Abstract read 2026-09-20. The canonical requirement experiment in this disease. Cited here as the exemplar of column one, not for any ordering or determinant content, which it does not have.
Evidence that argues against this (1)
- Requirement plus disease-selectivity is stronger than bare requirement and is what this paper reports - so column one is not uninformative, and the point of this entry is proportion rather than dismissal. Bach C et al. 2009
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Degron-mediated MLL::AF9 removal changes transcriptional elongation at core target genes within 15 minutes, with loss of the active chromatin landscape occurring subsequently.
MLL::AF9 degradation induces rapid changes in transcriptional elongation and subsequent loss of an active chromatin landscape. · Olsen SN, Godfrey L, Healy JP, Choi YA, et al. · Mol Cell 2022What this evidence cannot show
Abstract read 2026-09-20. The exemplar of column two, and in this disease it is close to unique: one fusion, one pair of events. It is cited here for the design rather than for the biology.
Evidence that argues against this (1)
- An ordering established by complete degradation does not transfer to partial pharmacological occupancy, so even a column-two result carries a scope limit that a reader can overstate in exactly the same way. Mahdavi L et al. 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Fusion proteins built on Mll2/KMT2B cannot transform haematopoietic cells, and domain swaps identify the CxxC and post-CxxC region rather than the shared menin-binding motif as the determinant.
Alterations of the CxxC domain preclude oncogenic activation of mixed-lineage leukemia 2. · Bach C, Mueller D, Buhl S, Garcia-Cuellar MP, et al. · Oncogene 2009What this evidence cannot show
Abstract read 2026-09-20; 'Mll2' is KMT2B per brief rule 66. The exemplar of column three: a comparator that holds the requirement constant and loses the phenotype. Engineered constructs in murine cells, one assay system.
Evidence that argues against this (1)
- KMT2B does sustain established MLL-rearranged AML in trans, so a paralogue that fails as a fusion backbone is not therefore irrelevant - a determinant experiment answers a narrow question too. Chen Y et al. 2017
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Mll-AF9 knock-in mice with or without Hoxa9 develop myeloid leukaemia with similar incidence and time course, and no single Hoxa gene among Hoxa5-Hoxa10 is necessary.
Hoxa9 influences the phenotype but not the incidence of Mll-AF9 fusion gene leukemia. · Kumar AR, Hudson WA, Chen W, Nishiuchi R, et al. · Blood 2004What this evidence cannot show
Abstract read 2026-09-20. The exemplar of column four's absence: without a dose axis, a buffered module looks either essential or dispensable depending on which experiment is quoted. Germline knockout, so compensation cannot be excluded.
Evidence that argues against this (1)
- In human KMT2A-rearranged lines HOXA9 depletion kills selectively, so the mouse result is not a licence to call HOXA9 dispensable; the two results are compatible and the compatibility is the point of this entry. Faber J et al. 2009
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In the phase 2 KMT2A-rearranged portion of AUGMENT-101, revumenib produced CR+CRh in 22.8% of 57 efficacy-evaluable patients with an overall response rate of 63.2%.
Menin Inhibition With Revumenib for KMT2A-Rearranged Relapsed or Refractory Acute Leukemia (AUGMENT-101). · Issa GC, Aldoss I, Thirman MJ, DiPersio J, et al. · J Clin Oncol 2025What this evidence cannot show
Abstract read 2026-09-20. Cited for the inference pattern only: a working drug makes its target look like a determinant, and clinical efficacy is a requirement result at the level of the patient.
Evidence that argues against this (1)
- Clinical resistance through MEN1 interface mutations shows the target is genuinely engaged and genuinely load-bearing, which makes the slide from requirement to determinant even more tempting and no more valid. Perner F et al. 2023
- E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
HOXA9 depletion causes proliferation arrest and apoptosis specifically in KMT2A-rearranged human lines and preferentially in KMT2A-rearranged primary AMLs.
HOXA9 is required for survival in human MLL-rearranged acute leukemias. · Faber J, Krivtsov AV, Stubbs MC, Wright R, et al. · Blood 2009What this evidence cannot show
Abstract read 2026-09-20. Paired with E4 to show two correct requirement experiments giving opposite headlines because they ask about different species and different disease stages.
Evidence that argues against this (1)
- shRNA depletion is partial, so this result also cannot supply a dose axis, and the apparent conflict with the mouse knockout may be partly a difference in how much protein each experiment actually removed. Kumar AR et al. 2004
same family of claim inflation: a scoped result quoted as an unscoped one
the distinction the chain's measured-versus-inferred table is built on
menin is the worked example of a requirement read as a determinant
the rate-limiting column, where no dose axis exists
both propose the same schema repair - record the design and the system on the evidence item