No proximity, degron or tiling-CRISPR study has ever been run on WRAD in KMT2A-fusion leukaemia
also called the modern-methods gap for WRAD · WRAD has never met a degron
Every claim about whether WRAD touches the leukemia fusion comes from methods that were old in 2004. None of the techniques that would actually settle it - proximity tagging, rapid protein degradation, fine-grained CRISPR - has ever been pointed at this question.
The same thing, for a scientist
A bounded absence, stated with its exact queries so it can be re-run. **Proximity labelling:** `(WDR5[tiab] OR RBBP5[tiab] OR ASH2L[tiab] OR DPY30[tiab]) AND (BioID[tiab] OR TurboID[tiab] OR "proximity labeling"[tiab] OR "proximity labelling"[tiab] OR APEX2[tiab])` returns **0** records [E1]. **Degron:** the same subunits against `(degron OR dTAG OR "auxin-inducible" OR PROTAC OR degrader)[tiab]` return 15 records, all WDR5 degrader chemistry and none in a KMT2A-fusion-association design [E1]. **CRISPR:** `WDR5[tiab] AND (CRISPR[tiab] OR sgRNA[tiab] OR "knockout screen"[tiab])` returns 4, and no tiling screen of any WRAD subunit in a KMT2A-rearranged line exists [E1]. The absence is real and it is **weaker than it looks**: a BioID study of the NUP98-PHF23 oncofusion recovering WDR5 and RBBP5 is invisible to the proximity query because its abstract names the methyltransferases at complex level [E2]. So the correct statement is not 'proximity labelling has never recovered WRAD around a fusion' - it has, in a sister family - but '**no proximity, degron or tiling-CRISPR experiment has been designed to ask whether WRAD reaches a KMT2A fusion**'.
Why this absence is the actionable one
The three missing method classes map one-to-one onto the three competing models this beat could not distinguish.
| Model | The method that would decide it |
|---|---|
| direct contact with the fusion | stringent co-IP of a tagged fusion, or a structure |
| second-shell passenger on wild-type KMT2A | proximity labelling with a wild-type-allele degron |
| no route at all | both of the above returning negative |
And a fourth question - whether a WRAD subunit has a non-catalytic, domain-localised requirement invisible to genome-wide screening - is exactly what a tiling CRISPR screen answers. That design has been run for BOD1L in two KMT2A-rearranged myeloid lines, MOLM-13 and MV4-11 (MV4-11 carries KMT2A::AFF1), and found a strong dependency on one domain in a gene DepMap classifies as non-essential in all leukaemia cell lines 3. The same design has never been pointed at WDR5, RBBP5, ASH2L or DPY30.
How to re-run this absence honestly
Any future cycle re-testing these counts should (a) run the subunit queries and complex-name phrases ("MLL core complex", COMPASS, "KMT2/MLL", "H3K4 methyltransferase complex"), because the subunit-only version demonstrably leaks; (b) avoid any query containing a bare der(n), which PubMed parses as syntax and silently zeroes; (c) treat a zero as a statement about indexing. The numbers here are a floor on what exists, not a ceiling.
| proximity records | 0 |
|---|---|
| degron records | 15 |
| crispr records | 4 |
| run date | 2026-09-19 |
| known leak | complex-level framing; PMID 39922194 recovers WDR5 and RBBP5 by BioID around an oncofusion and returns zero from the proximity query |
| absence class | never-measured |
| absence class basis | None of the three designs has been applied to WRAD in this disease. |
| absence class assigned | 2026-09-21 (cycle 47, king) |
queries
{
"proximity": "(WDR5[tiab] OR RBBP5[tiab] OR ASH2L[tiab] OR DPY30[tiab]) AND (BioID[tiab] OR TurboID[tiab] OR \"proximity labeling\"[tiab] OR \"proximity labelling\"[tiab] OR APEX2[tiab])",
"degron": "(WDR5[tiab] OR RBBP5[tiab] OR ASH2L[tiab] OR DPY30[tiab]) AND (degron[tiab] OR dTAG[tiab] OR \"auxin-inducible\"[tiab] OR PROTAC[tiab] OR degrader[tiab])",
"crispr": "WDR5[tiab] AND (CRISPR[tiab] OR sgRNA[tiab] OR \"knockout screen\"[tiab])"
}feasible now
[ "tiling CRISPR of ASH2L and DPY30 in MV4-11 or SEM", "TurboID from a tagged KMT2A::AFF1 with a wild-type-allele degron" ]
- Stating that a question has never been asked can be read as an invitation to ask it, and the four proteins involved are shared across six complexes with documented marrow toxicity - so running the experiment is cheap and acting on a positive would not be.
- A degron experiment on the wild-type KMT2A allele removes a protein the leukaemia cell requires for maintenance, so the viability window may be too short to read any result in.
- Counted absences in this graph have twice produced confident wrong numbers; this one is published with its own falsifying counterexample attached precisely so it is not read as stronger than it is.
- Does a tiling CRISPR screen of ASH2L or DPY30 in a KMT2A::AFF1 line reveal a domain-localised requirement? Neither is pan-essential, so the design is straightforwardly feasible and has never been run.
- Is there an unpublished proximity dataset with a KMT2A-fusion bait? This project has previously found relevant depositions through elink from PMID to GEO where the paper itself did not state them.
- What is the right way to write a counted absence in a field that frames results at complex level? Every absence claim in this beat inherits the leak demonstrated here.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Title/abstract PubMed queries return 0 records pairing any WRAD subunit with any proximity-labelling method, 15 pairing them with degradation methods (all WDR5 degrader chemistry, none in a fusion-association design), and 4 pairing WDR5 with CRISPR screening terms.
PubMed E-utilities esearch, method-class queries run 2026-09-19; exact query strings in this entry's summary and in ana-wrad-fusion-evidence-enumeration · NCBI PubMed; queries run by this project · Database 2026What this evidence cannot show
Title/abstract restriction is deliberate - it is what makes the query reproducible - and is also why it under-counts. CRISPR maps to a restrictive MeSH term in untagged queries, which is why the tagged form was used. Supplementary data are invisible.
Evidence that argues against this (1)
- The NUP98-PHF23 BioID study recovers WDR5 and RBBP5 and returns zero from this query, so the proximity absence is an indexing fact rather than a literature fact. Ahn JH et al. 2025
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
BioID proximity labelling of the NUP98-PHF23 oncofusion recovers KMT2A/MLL1, KMT2B, WDR5 and RBBP5, demonstrating both that the method can recover WRAD around a fusion oncoprotein and that such a result is unreachable from any WRAD-keyed literature query.
The phenylalanine-and-glycine repeats of NUP98 oncofusions form condensates that selectively partition transcriptional coactivators. · Ahn JH, Guo Y, Lyons H, et al. · Mol Cell 2025What this evidence cannot show
A different fusion family with no KMT2A sequence in the fusion. A proximity hit is a neighbour, not a binding partner. The subunit identities are in the paper's data rather than its abstract.
Evidence that argues against this (1)
- An unbiased AP-MS interactome of five NUP98 fusions at near-physiological levels explicitly failed to detect the reported MLL1 and NSL complex interactions, so even the precedent is contested at the level of the primary assay. Terlecki-Zaniewicz S et al. 2021
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
BOD1L is classified as non-essential in all leukaemia cell lines in DepMap, yet a CRISPR-tiling screen in doxycycline-inducible Cas9 MOLM-13 and MV4-11 human leukaemia lines identifies a strong dependency on the N-terminal Shg1 domain and not on the C-terminal disordered regions or AT-hook; BOD1L loss reduces SETD1A occupancy at transcription start sites and causes an elongation defect without reducing H3K4me3 - a non-catalytic SET1-family requirement that genome-wide screening cannot see.
BOD1L mediates chromatin binding and non-canonical function of H3K4 methyltransferase SETD1A. · Hoshii T, Kikuchi S, Kujirai T, et al. · Nucleic Acids Res 2024What this evidence cannot show
Full text read via PMC11381347. The screen lines are MOLM-13 and MV4-11, both KMT2A-rearranged myeloid lines; MV4-11 carries KMT2A::AFF1. The target is SETD1A-adjacent biology, not WRAD. Tiling screens have limited dynamic range in genes that are already pan-essential, which WDR5 and RBBP5 are, so the design may not transfer to all four subunits.
Evidence that argues against this (1)
- ASH2L and DPY30 are NOT pan-essential (21.4% and 31.5% of lines depleted), so for those two the dynamic-range objection does not apply and the screen is straightforwardly feasible. Broad Institute DepMap; Open Targets Platform 2026
the same hole seen from the cell-state side
the four WRAD same-source absence notes were checked against this bounded absence and stand; the complex form was run, so Rule 64 is satisfied
SURVIVES for the proximity and tiling arms after a 74-fold recall increase; the degron arm is an unscreened upper bound of 92
SURVIVES a WDR5-degrader-name and leukaemia-cell-line re-query
SURVIVES: MS67 is a 3.7 nM chemical degron in MV4;11 whose MLL blot cannot distinguish the fusion
four hours of MS67 and two KMT2A antibodies would close this absence's degron arm
The method-class absences are the sharpest counted result in the enumeration.
- related to →A BioID hit is a neighbour, not a complex member - and the two methods are known to disagree
No proximity-labelling study has ever been designed around a WRAD subunit and a KMT2A fusion; the concept entry states what such a study could and could not show.
The tiling-CRISPR design that found a domain-localised dependency in MV4-11 exists, works in a KMT2A::AFF1 line, and has never been applied to a WRAD subunit.
- supports →Settle the WRAD question: tagged co-IP plus degron-timed proximity across three fusion classes
The absence names exactly the experiment this entry designs.