Scored against the literature: the composition half of the YEATS-selectivity objection holds and the selectivity half is refuted — a scaffold-blind probe already has an in vivo window in the index fusion with no HSPC toxicity
also called does the YEATS scaffold-blindness argument survive? · q187 verdict on reader selectivity
The worry was that because the two mark-reading proteins sit in three different machines, a drug aimed at them cannot be specific to the leukaemia. The composition part of that is true. The conclusion is not: somebody already ran the drug in mice with this exact leukaemia and it killed the cancer without harming blood stem cells.
The same thing, for a scientist
The question this beat was dispatched with was: *ENL and AF9 are in all three SEC-family complexes, which control different gene sets, so a YEATS inhibitor cannot be fusion-selective on the basis of where the readers are — is that supported?* **Scored in two halves, the halves go opposite ways.** The compositional half is supported and is close to a tautology once the purification is granted: the YEATS pocket is not the scaffold-binding surface, so no ligand in it can discriminate between SEC, SEC-L2 and SEC-L3 by scaffold [E1][E2]. The selectivity half — that a YEATS inhibitor *therefore* cannot be fusion-selective — is **refuted by an existing measurement this corpus already held and this beat re-read**: SGC-iMLLT downregulated MLLT1 target genes and strongly inhibited KMT2A::AFF1-driven leukaemia growth in vitro and in vivo while not altering human HSPC colony-forming potential or long-term mouse HSPC function in vivo [E3]. A therapeutic window exists for a compositionally scaffold-blind agent in the index fusion. **The lesson is a general one about this graph's reasoning: a selectivity argument built on where a target protein resides cannot be settled by composition, because a window can be produced by differential dependence with no molecular selectivity at all** — which is precisely how menin inhibitors work in this same disease, against a protein expressed in every tissue. The composition result is not worthless: it **relocates** the liability from haematopoietic tissue, where it has been measured and cleared, to imprinting, X-inactivation and neurodevelopment, where it has never been looked for.
The two halves, scored
| claim | verdict | what settles it |
|---|---|---|
| ENL and AF9 are subunits of SEC, SEC-L2 and SEC-L3 | holds | MudPIT + endogenous co-IP + Superose 6 in the defining purification 1 |
| The three complexes control largely different gene sets | holds, quantified | 1,539-gene union, 27.5% invisible to AFF4 depletion (ana-q187-sec-family-target-set-overlap-measured) 1 |
| A YEATS-pocket ligand cannot discriminate between them by scaffold | holds | the AHD, not the YEATS domain, is the scaffold-binding surface 2 |
| Therefore a YEATS inhibitor cannot be fusion-selective | REFUTED | SGC-iMLLT: efficacy in KMT2A::AFF1 in vivo, human HSPC colonies and long-term mouse HSPC function unaffected 3 |
The fourth row does not follow from the first three, and the reason is worth stating plainly because this graph makes the same move elsewhere: the first three are about where the drug binds and the fourth is about what happens to a cell. Those are connected by dependence, not by residence.
Why a composition argument cannot decide selectivity — three routes, all live here
1. A window can come from dependence with no molecular selectivity at all. This project's own flagship comparator proves it: menin is expressed in essentially every tissue and menin inhibitors have a clinical window in KMT2A-rearranged leukaemia. The window is produced by the fusion cell being closer to the cliff, not by the drug distinguishing leukaemic menin from normal menin. Any argument of the form 'the target is present in normal tissue, therefore no window' is refuted inside this graph before it is applied to YEATS.
2. The reader pool may be lopsided and nobody has measured it. AFF2 and AFF3 mRNA are roughly 10-fold lower than AFF4 in the one cell background where the three were compared 1. If most ENL and AF9 is in SEC, the sibling complexes absorb a small share of pocket occupancy, and scaffold-blindness is a true statement with a negligible quantitative consequence. No experiment in any system reports the partition of the ENL or AF9 pool across SEC, SEC-L2, SEC-L3, DotCom and free protein. That single unmeasured number controls the size of the whole effect.
3. Membership is not requirement. Wan and colleagues found ENL required for AML maintenance and its paralogue AF9 not required — two proteins in the same three complexes, one load-bearing and one not 2. If co-membership does not equalise ENL and AF9, it does not establish that the reader is load-bearing in SEC-L2 or SEC-L3 either. Applied consistently, the argument weakens itself.
What the composition result DOES buy, which is a relocation rather than a refutation
The published window 3 is measured against haematopoietic endpoints: an in vitro human HSPC colony-forming assay and a long-term in vivo mouse HSPC function arm. Those are the right endpoints for the liability the field expected — ENL is expressed in whole blood at about 40 TPM and MLLT3/AF9 governs human HSC self-renewal (target-phase2-enl-yeats-dossier).
They cannot see the liability the composition result points at. protein-sec-l3-complex records AFF3's documented physiological function as allele-specific regulation at gametic DMRs, the Meg3 enhancer and the XIST DMR; protein-sec-l2-complex records AFF2 as the FRAXE gene. Neither a colony assay nor a transplant arm reports allele-specific expression, imprinting status or neurodevelopment, and the patients are infants.
So the correct summary is not 'the objection was wrong'. It is:
The objection was directed at the wrong endpoint. As a claim about fusion selectivity it is refuted. As a claim about where the unmeasured risk is, it survives intact and is the most actionable thing this beat produced, because the haematopoietic clearance is now published and the non-haematopoietic question has never been asked.
That is recorded as risk-q187-yeats-inhibition-touches-an-imprinting-complex-in-an-infant and absence-q187-no-yeats-inhibitor-has-been-tested-against-sec-l2-or-sec-l3.
A correction to report on another beat's entry (not edited here)
mechanism-q181-aff2-and-aff3-carry-the-same-yeats-readers-as-the-index-fusion-partner (beat q181, cycle 159) states in bite_back[0]:
"If those readers are also stoichiometric members of SEC-L2 and SEC-L3 in normal tissue, a YEATS inhibitor is not a fusion-selective agent — it disables the reader module of three elongation complexes at once…"
The conditional protects most of it, and the second clause is correct. The first clause — 'is not a fusion-selective agent' — is contradicted by PMID 42242459 (Rajhansa et al., Exp Hematol 2026;105458, DOI 10.1016/j.exphem.2026.105458), abstract read 2026-09-24 via efetch; full text not in PMC and not read. That paper reports efficacy of SGC-iMLLT in KMT2A::AFF1 in vitro and in vivo with human HSPC colony formation and long-term mouse HSPC function unaffected, which is a measured therapeutic window for exactly the class of agent the sentence says cannot be selective.
Suggested amendment, for the owning beat to make: replace 'a YEATS inhibitor is not a fusion-selective agent' with 'a YEATS inhibitor cannot be selective BY SCAFFOLD COMPOSITION; whether it is selective in effect is a dependence question, and PMID 42242459 reports a window in KMT2A::AFF1 with haematopoietic endpoints clear'. The rest of that bite-back — that the reader module of three complexes is disabled at once, and that the target sets are distinct and uncharacterised — stands and is strengthened by this beat's numbers.
| status | run |
|---|---|
| method | claim decomposition and scoring against fetched primary literature; no computation |
| result summary | 3 of 4 component claims hold; the conclusion claim is refuted by PMID 42242459. Verdict: composition cannot settle selectivity; the composition result relocates the unmeasured liability from haematopoietic to imprinting/neurodevelopmental tissue. |
| third option preregistered | Before scoring, the possible verdicts were written as three, not two: SUPPORTED, REFUTED, or NEITHER — the argument is about a question composition cannot answer. The answer landed on the third for the conclusion clause and on SUPPORTED for the three component clauses, which is why this entry reports a split verdict rather than a yes or a no. |
inputs
[ "PMID 22547686 (full text, PMC3434493 HTML)", "PMID 28241141 (abstract)", "PMID 23260655 (abstract)", "PMID 42242459 (abstract)", "PMID 30288907 (abstract)", "existing corpus entries drug-sgc-imllt, target-phase2-enl-yeats-dossier, mechanism-q181-aff2-and-aff3-carry-the-same-yeats-readers-as-the-index-fusion-partner" ]
correction reported not applied
{
"entry": "mechanism-q181-aff2-and-aff3-carry-the-same-yeats-readers-as-the-index-fusion-partner",
"field": "bite_back[0]",
"text_at_issue": "a YEATS inhibitor is not a fusion-selective agent",
"should_say": "a YEATS inhibitor cannot be selective by scaffold composition; whether it is selective in effect is a dependence question, and PMID 42242459 reports an in vivo window in KMT2A::AFF1 with haematopoietic endpoints clear",
"citation_that_settles_it": "PMID 42242459, DOI 10.1016/j.exphem.2026.105458",
"read": "abstract only, via efetch 2026-09-24; full text not in PMC",
"note": "Reported, not edited — the entry belongs to beat q181."
}measurements
[
{
"quantity": "component_claims_upheld",
"subject": "the four-clause YEATS scaffold-blindness argument as stated in the q187 dispatch brief",
"value": 3,
"units": "clauses of 4",
"n": 4,
"universe": "clauses scored individually against fetched primary sources: complex co-membership, distinct target sets, surface separation, and the fusion-selectivity conclusion",
"null": null,
"note": "The one refuted clause is the conclusion, which is the clause that would be quoted. A 3-of-4 score here is not 75% right; it is right about the premises and wrong about the inference."
}
]- This entry argues that a composition-based objection was too strong. The symmetric error is to now treat the window as settled: it is one study, abstract-only to this beat, from a group overlapping the probe's originators, with no on-target mutant rescue and no non-haematopoietic endpoint. Downgrading the objection should not upgrade the drug.
- The dual MLLT1/MLLT3 probe hits the paralogue that governs human HSC self-renewal as hard as the one that is required in leukaemia. A clean HSPC colony result is the least sensitive possible way to look for a self-renewal defect; the mouse long-term arm is the load-bearing one and it is a single experiment in a single paper.
- If the correct reading is 'selectivity comes from dependence, not composition', then the right question for this target is how steep the dependence is in KMT2A::AFF1 versus normal progenitors — and `target-phase2-enl-yeats-dossier` records the MEN1 co-dependency as weak when pooled (r 0.130) and only rising to 0.426 inside leukaemia lines, which is a subgroup correlation computed after the pooled one came out near zero. The dependence case is not strong enough to carry the weight this entry has just moved onto it.
- What is the reader-pool partition across SEC, SEC-L2, SEC-L3, DotCom and free protein in a KMT2A::AFF1 line? Every quantitative form of the scaffold-blindness argument needs this number and it does not exist.
- Does the SGC-iMLLT window survive a non-haematopoietic readout? Allele-specific expression at an imprinted locus in a treated human cell line is a cheap first look and nobody has taken it.
- Is the in vivo efficacy in PMID 42242459 on-target? No YEATS aromatic-cage mutant rescue has been run in any KMT2A::AFF1 line, so an off-target contribution to the window has not been excluded — and an off-target window is not a window.
- How often does this graph make the same move — inferring a selectivity verdict from where a protein resides? The menin counter-example is in this corpus already, so the pattern is detectable; a sweep of selectivity claims for residence-based reasoning would be a one-cycle audit.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
ENL/MLLT1 and AF9/MLLT3 are subunits of all three AFF-scaffolded P-TEFb complexes (SEC, SEC-L2, SEC-L3), the three control largely different gene subsets, and AFF2 and AFF3 mRNA are about 10-fold lower than AFF4 in the HEK293T background used.
The super elongation complex family of RNA polymerase II elongation factors: gene target specificity and transcriptional output. · Luo Z, Lin C, Guest E, Garrett AS, Mohaghegh N, Swanson S, Marshall S, Florens L, Washburn MP, Shilatifard A · Molecular and Cellular Biology 2012What this evidence cannot show
HEK293/293T only, over-expressed tagged baits, one laboratory, 2012. No haematopoietic cell, no leukaemia line, no fusion protein anywhere in the paper. The gene-set comparison is a knockdown-response comparison, not an occupancy comparison.
Evidence that argues against this (1)
- Lu and colleagues argue the SEC family is combinatorial rather than a set of three fixed particles — different ELL and AFF content serving different genes, with HIV-1 Tat selecting among species. Under that model a three-complex table overstates the tidiness of the object and the scaffold-blindness argument becomes harder to state, not easier: there is no clean set of three things for a ligand to fail to distinguish between. Their antibody panel does not include AFF2 or AFF3, so they neither confirm nor contradict SEC-L2 and SEC-L3 composition. Lu H et al. 2015
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The YEATS domain and the ANC1 homology domain are separate surfaces with separate jobs — the YEATS domain binds acetylated histone H3 and its structure-based disruption reduces Pol II recruitment without removing the protein, while the AHD is the intrinsically disordered surface that recruits AFF-family scaffolds — and in AML ENL is required for disease maintenance while its paralogue AF9, resident in the same complexes, is not.
ENL links histone acetylation to oncogenic gene expression in acute myeloid leukaemia. · Wan L, Wen H, Li Y, Lyu J, Xi Y, Hoshii T, Joseph JK, Wang X, Loh YE, Erb MA, Souza AL, Bradner JE, Shen L, Li W, Li H, Allis CD, Armstrong SA, Shi X · Nature 2017What this evidence cannot show
Myeloid AML, not infant B-lineage KMT2A::AFF1. The AHD half of this claim is carried by a separate paper (PMID 23260655, AF9 AHD-AF4 NMR structure) and is cited as E1 of `mechanism-q187-target-specificity-is-a-scaffold-property-while-the-reader-is-shared`; this item states the YEATS half and the ENL-versus-AF9 asymmetry, both of which are in this abstract. A genetic mutant is not a ligand.
Evidence that argues against this (1)
- The ENL-required / AF9-not-required asymmetry is cited here to show that co-membership does not imply co-importance, which weakens the composition argument. It equally undermines a reader-directed strategy's simplicity in the other direction: the available chemical probe, SGC-iMLLT, does NOT distinguish MLLT1 from MLLT3, so a probe hits the dispensable paralogue as hard as the required one and MLLT3/AF9 is the one that governs human HSC self-renewal. Moustakim M et al. 2018
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
SGC-iMLLT, a compositionally scaffold-blind YEATS-pocket probe, downregulated MLLT1 target genes and strongly inhibited KMT2A::AFF1-driven leukaemia growth both in vitro and in vivo, while not altering the in vitro colony-forming potential of human HSPCs or the long-term in vivo function of mouse HSPCs — so the inference 'the readers are everywhere, therefore no fusion selectivity' is contradicted by a direct measurement in the index fusion.
Inhibition of MLLT1 Limits Growth of KMT2A::AFF1 Leukemias Without Killing Healthy Hematopoietic Stem Cells. · Rajhansa S, Crump NT, Khoo HM, Dopico-Fernandez A, Bozhilov Y, Brennan PE, Fedorov O, Adams C, Farnie G, Milne TA, Wilkinson AC · Experimental Hematology 2026What this evidence cannot show
Abstract only; the full text is not in PMC and this beat did not read it. Unknown: the cell-line panel, concentrations, schedule, the in vivo model, and what 'long-term in vivo function' was measured as. One study, one group, and the group overlaps with the probe's originating consortium. No YEATS aromatic-cage mutant rescue has been run in a KMT2A::AFF1 line, so the on-target basis of the in vivo efficacy is assumed rather than shown — `drug-sgc-imllt` records that gap as its own open question. KMT2A::AFF1 carries no YEATS domain, so the efficacy is wild-type-ENL pharmacology and is not obviously fusion-stratified biology.
Evidence that argues against this (1)
- The window is measured only against haematopoietic endpoints and therefore cannot address the liability the composition result actually points at. AFF3, the SEC-L3 scaffold, binds the DMR downstream of the XIST promoter in a DNA-methylation-dependent way and its knockdown de-represses the inactive XIST allele in terminally differentiated cells; nothing in a colony assay or an HSPC transplant arm would detect an equivalent effect. A clean haematopoietic safety result is fully compatible with an undetected allele-specific-expression liability. Zhang Y et al. 2019
upholds the composition clauses and refutes the fusion-selectivity conclusion drawn from them
bite_back[0]'s clause 'a YEATS inhibitor is not a fusion-selective agent' is contradicted by PMID 42242459; correction reported, not applied
the probe and the 2026 in vivo window this scoring turns on
moves the selectivity case onto dependence, which the dossier records as weak when pooled (r 0.130)
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
the index fusion is the system in which the window was measured
what survives of the objection after the selectivity clause is refuted: the liability is relocated, not removed
refutes the unqualified selectivity claim and leaves the partition question as the testable residue