MED26 — a common-essential Mediator subunit with a solved 87-residue domain and no ChEMBL target entity at all

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

• Essential-function toxicity with no compensating selectivity. MED26 is common-essential with 92.3% of 1,208 lines depleted, and its KMT2A::AFF1 mean (-0.718) is the shallowest of the three cancer groupings — so the fusion of interest is, if anything, the least MED26-dependent group. Any agent would carry general-transcription toxicity and buy nothing genotype-specific.
• The central mechanistic rationale for including MED26 here is unverified. The Mediator-to-SEC handoff is our inference, labelled speculative (E4), and it is the only thing that would make MED26 interesting for this fusion. If it is wrong, MED26 reduces to a generic essential Mediator subunit and should be dropped from the target list rather than merely deprioritised.
• Compensation at the complex level is likely and unmeasured. Mediator is a ~30-subunit modular assembly; losing one peripheral subunit can be partly absorbed by the rest, which is consistent with MED26's -0.944 being shallower than CDK9's -1.536 despite both being common-essential. A partial inhibitor would sit in exactly the regime that Mediator buffers best, giving a flat dose-response until the toxic dose.
• Delivery and modality are unsolved from the ground up. There is no ligand, no ligand-bound structure, no assay in ChEMBL, and no chemical series — so a MED26 programme would begin by establishing ligandability on a domain that has never been screened, on a target with a disqualifying essentiality profile. The cheap version of that experiment is worth running for knowledge; the expensive version is not worth running for this disease.
• Resistance would be cheap for the tumour and expensive for the patient: a common-essential target with no selectivity means the therapeutic dose is close to the toxic dose, so the achievable exposure is low, sub-maximal target engagement is the norm, and sub-maximal engagement of a modular complex is the classic setting for rapid adaptive escape.

1. Is the MED26 TND actually a partner-exchange switch, and does anything bind it competitively? Both solved structures (5ODD, 6ZV3) are apo with no partner chain, so the surface's real ligand has never been visualised. A TND-deleted or TND-point-mutant knock-in, reading out Mediator versus SEC co-immunoprecipitation and nascent transcription at KMT2A::AFF1 target genes, would decide it — and would distinguish a genuine switch from a merely weaker Mediator subunit.
2. Why has nobody screened the MED26 TND? It is the same size class as the CBX8 chromodomain, the BRPF1 bromodomain and the ENL/AF9 YEATS domains, all of which have tens to hundreds of potent ligands, and it is more essential than any of them. Whether the groove is too shallow, or the field simply wrote Mediator off, is unknown — and a 10,000-compound thermal-shift screen on a 10 kDa NMR-tractable domain would answer it cheaply.
3. What does MED26 residues 93-600 look like? About 86% of the protein has never been resolved outside a holo-assembly, so there may be a second autonomous module nobody has looked for. AlphaFold AF-O95402-F1 exists and would be the place to start, but a prediction is not a structure.
4. Does MED26's selectivity sign-flip (-0.113 against leukemia, +0.031 against all lines) survive a larger KMT2A-r panel, or is it purely n=10 artefact? Given that the KMT2A::AFF1 subgroup is the least depleted grouping of all, the honest prior is that MED26 has no fusion-specific requirement at all.
5. MED26 shows tight LoF constraint (LOEUF 0.301, 7 observed of 43.7 expected) with essentially no missense constraint (z 0.83). Does that mean MED26 is inhibitable only by degradation, because no point change in it is deleterious enough to model what an occupancy-based inhibitor would do?

1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   MED26 is DepMap common-essential with a mean Chronos of -0.944 across 1,208 cell lines and 92.3% of lines depleted; means are -0.800 across 64 leukemia lines, -0.913 across 10 KMT2A-rearranged lines and -0.718 across 4 KMT2A::AFF1 lines, giving selectivity_vs_leukemia -0.113 and selectivity_vs_all_lines +0.031 — a sign-unstable score, both magnitudes inside noise. gnomAD gives pLI 1.0, LOEUF 0.301, 7 observed versus 43.7 expected loss-of-function variants and missense z 0.83. HPA classifies the RNA as tissue enhanced (testis, 34.0 nTPM), detected in all tissues, with low immune cell specificity and blood-cell distribution 'detected in single'.

   DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL API; loss-of-function constraint via the gnomAD GraphQL API; normal-tissue expression via the Human Protein Atlas API · Broad Institute DepMap; gnomAD; Human Protein Atlas; Open Targets Platform · Databases 2026
   ▸What this evidence cannot show

   Means were computed by us from per-line API values rather than read from a DepMap release table, and the API does not state the release version. Only 10 KMT2A-rearranged lines (4 KMT2A::AFF1) contribute; the KMT2A::AFF1 group spans -0.358 (MV4-11) to -0.881 (SEM), so its mean of -0.718 carries little weight, and the fact that selectivity changes sign between the two comparators (-0.113 vs +0.031) is itself evidence that neither is resolvable at this n. Cell lines are not normal tissue. Whole-gene knockout removes the TND and the remaining 513 residues together and so cannot speak to what blocking the TND alone would do. 'Detected in single' describes RNA detection across immune cell types, not protein requirement.

   ▸Evidence that argues against this (1)

   • The KMT2A::AFF1 lines are the LEAST depleted of the three cancer groupings (-0.718 vs -0.913 for all KMT2A-r and -0.800 for leukemia), so if anything the fusion of interest is the subgroup with the weakest MED26 requirement. Combined with the +0.031 selectivity against all lines, there is no dependency argument for MED26 in this disease. Broad Institute DepMap; Open Targets Platform 2026
2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   MED26's structural coverage is bimodal with nothing between the extremes: the isolated N-terminal domain has been solved twice by NMR — 5ODD (MED26 residues 1-92, N-terminal domain, associated with PMID 26861138) and 6ZV3 (MED26 residues 1-87, annotated as the TFIIS N-terminal domain, associated with PMID 34822292) — while MED26 otherwise appears only as one subunit within large pre-initiation-complex–Mediator cryo-EM assemblies (7ENA, 7ENC, 7EMF, 7ENJ, associated with PMID 33958484; 8GXQ, 8GXS, associated with PMID 36201575). No structure exists of a MED26 fragment bound to a partner peptide, no sub-module structure exists, and residues 93-600 have never been resolved outside a holo-complex.

   RCSB PDB entries 5ODD, 6ZV3, 7ENA, 7ENC, 7EMF, 7ENJ, 8GXQ and 8GXS, with per-entity UniProt accessions, methods and modelled residue ranges resolved against the RCSB APIs; searched for MED26 (O95402) complexes with partner peptides · RCSB Protein Data Bank · RCSB PDB 2026
   ▸What this evidence cannot show

   Residue ranges are RCSB's UniProt-mapped modelled segments, not independently curated domain boundaries. Both TND structures are apo and isolated: they contain no ligand and no partner chain, which is the least useful starting point for designing a PPI blocker. Within the PIC-Mediator assemblies the stated global resolution does not apply to a peripheral subunit such as MED26, whose local resolution is generally worse and was not independently assessed here. The claim that no partner-bound MED26 fragment structure exists is scoped to what an RCSB API search returned in this session. BOUND ADDED 2026-09-22, and the enumeration is EXHAUSTIVE. The RCSB search API, queried on the polymer-entity reference accession for O95402, returns exactly 8 PDB entries: 5ODD, 6ZV3, 7EMF, 7ENA, 7ENC, 7ENJ, 8GXQ, 8GXS. This claim names all eight, so 'MED26 otherwise appears only as one subunit within large PIC-Mediator assemblies' is a complete statement of the deposited record on that date and not a sample of it. The count will drift as new entries are deposited -- the same caveat gene-kmt2a's structural-coverage evidence carries -- so the date is the load-bearing part of this sentence.

   ▸Evidence that argues against this (1)

   • Counting PDB entries measures how tractable a protein was to CRYSTALLISE, which correlates with being small, ordered and well-behaved rather than with being druggable. What decides design value is whether the deposited density covers the site that matters, at a resolution that supports modelling, in the relevant state — and this graph has repeatedly found structures described in the direction that overstates what is resolved. Bimodal coverage with nothing between the extremes describes a protein whose middle is disordered, and two NMR structures of the same N-terminal domain is one construct solved twice. RCSB PDB 2026
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   MED26 has no ChEMBL target entity of any kind in ChEMBL v34 — neither a single-protein target with zero activity rows nor a protein-complex or protein-protein-interaction target — placing it with ZFP64 and MEAF6 as the only completely unrepresented genes in this interactome. By contrast, comparably sized autonomous interaction modules in the same dossier carry substantial chemical matter: the CBX8 chromodomain target CHEMBL3232684 has 15 potent rows and a 120 nM Kd, BRPF1 (CHEMBL3132741) has 173 potent rows and a 31 nM Kd, and the MLLT1 and MLLT3 YEATS targets (CHEMBL4295801, CHEMBL4295761) have 144 and 66 potent rows with best Ki values of 56 nM and 10 nM.

   ChEMBL database version 34: target_search for MED26 / Mediator complex subunit 26 returned no target entity; comparator targets CHEMBL3232684, CHEMBL3132741, CHEMBL4295801 and CHEMBL4295761 queried by target_search and get_bioactivity · EMBL-EBI ChEMBL · ChEMBL 2026
   ▸What this evidence cannot show

   Absence of a ChEMBL target is absence of curated records, not proof of intractability, and ChEMBL curation lags the primary and patent literature. It is equally consistent with the TND groove being too shallow to bind small molecules, with Mediator being perceived as undruggable and therefore unscreened, or with no screen having been attempted. The comparator counts are page-local because of a 25-row API limit. The comparison to CBX8, BRPF1 and the YEATS domains is an argument about module size class, not evidence that the MED26 TND is ligandable.

   ▸Evidence that argues against this (1)

   • A ChEMBL activity-row count measures how much medicinal chemistry has been DONE, not how tractable the target is, and the best reported affinity belongs to the most optimised compound in whichever series someone chose to pursue. Zero rows means nobody tried; thousands of rows means a target had a decade of industrial attention. No target entity of any kind is a statement about ChEMBL's curation policy as much as about the protein: ChEMBL creates targets when assays are deposited, so absence records that nobody has assayed it, which is the expected state for a Mediator subunit. ChEMBL, EMBL-EBI 2026
4. E4speculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.database

   OUR INFERENCE, NOT A VERIFIED FINDING: that the MED26 TND functions as the switch point handing RNA polymerase II from a Mediator-associated initiation state to a super-elongation-complex-associated elongation state. What is verifiable is only the structural furniture — that an autonomous TND exists and has been solved twice (5ODD, 6ZV3), and that MED26 is a subunit of PIC-Mediator assemblies (7ENA, 7ENC, 7EMF, 7ENJ, 8GXQ, 8GXS). No verified primary citation supporting the Mediator-to-SEC handoff function was obtained this cycle.

   RCSB PDB entries 5ODD, 6ZV3, 7ENA, 7ENC, 7EMF, 7ENJ, 8GXQ and 8GXS — cited here for the structural facts only; the functional switch interpretation is this project's inference and is not supported by these entries · RCSB Protein Data Bank · RCSB PDB 2026
   ▸What this evidence cannot show

   This evidence item exists to record an unsupported claim honestly rather than to support it. The structures establish that the TND is an autonomous module; they establish nothing about partner exchange, nothing about the SEC, and nothing about KMT2A fusions. Anyone citing this node for the handoff mechanism is citing an inference. The claim is developed with the same labelling, and with node-level status 'unverified', in interface-med26-sec-handoff.

   ▸Evidence that argues against this (1)

   • Both MED26 TND structures are apo with no partner chain modelled, so there is no structural evidence of competing partners occupying the same TND surface — which is what a switch mechanism requires. The absence of any partner-bound MED26 fragment structure in the PDB is a direct argument that the switch model has not been structurally tested. RCSB Protein Data Bank 2026
5. E_q77_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   MED26's ENTIRE subcellular record in UniProt O95402 (fetched 2026-09-21) is one line - 'Nucleus' - carrying ECO:0000305, which is curator inference, NOT experimental evidence. There is no ECO:0000269 anywhere in the block, no note, and no isoform- or fragment-scoped statement.

   UniProtKB O95402 (MED26_HUMAN) - SUBCELLULAR LOCATION comment block with evidence codes · UniProt Consortium 2026
   ▸What this evidence cannot show

   This is a measurement of the DATABASE, not of the protein, and it is reported as such. Fetched live 2026-09-21 by reading the comment array with evidence codes attached, not by reading the flattened cc_subcellular_location field, which hides the codes. The practical consequence is the one brief rule 91 item 4 warns about: a pipeline that keeps only ECO:0000269 rows returns nothing for MED26, and a pipeline that ignores codes returns 'nuclear, experimental', and both are wrong.

   ▸Evidence that argues against this (1)

   • An uncoded or curator-inferred annotation being absent from the database is NOT evidence that the measurement is absent from the world. For MED26 the experimental subnuclear record exists and is specific - Cajal bodies, with degron depletion and termination assays behind it - and it simply has not been curated into this record. Treating the ECO gap as a knowledge gap would have produced a confident wrong bounded absence. Takahashi H et al. 2020
6. E_q77_loc2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   MED26 recruits the little elongation complex (LEC) to modulate transcription TERMINATION of non-polyadenylated transcripts - snRNAs and replication-dependent histone mRNAs - AT CAJAL BODIES; the LEC recruited by MED26 promotes efficient Pol II termination through interaction with CBC-ARS2 and NELF/DSIF and promotes 3'-end processing by enhancing recruitment of Integrator to snRNA genes and Heat Labile Factor to replication-dependent histone genes.

   The role of Mediator and Little Elongation Complex in transcription termination. · Takahashi H, Ranjan A, Chen S, Suzuki H, Shibata M, Hirose T, Hirose H, Sasaki K, Abe R, Chen K, He Y, Zhang Y, Takigawa I, Tsukiyama T, Watanabe M, Fujii S, Iida M, Yamamoto J, Yamaguchi Y, Suzuki Y, Aoto Y, Tanaka H, Sugiyama S, Kanemaki MT, Ohkawa Y, Conaway JW, Conaway RC, Hatakeyama S · Nat Commun 2020
   PMID 32102997DOI
   ▸What this evidence cannot show

   Human cultured cells; no haematopoietic system and no fusion. Cajal bodies are named as the site of the process rather than as a steady-state address for MED26 protein, so the defensible statement is that MED26's LEC-recruiting function OCCURS at Cajal bodies, not that a given fraction of MED26 resides there - a distinction this entry keeps because the graph's condensate rules require it. NELF and DSIF are named in the mechanism and both are central to this project's elongation model, which is why this result is worth carrying despite the system mismatch.

   ▸Evidence that argues against this (1)

   • Four of the papers behind the Cajal-body model - the 2011 docking paper, the 2015 snRNA paper, this 2020 termination paper and the 2023 review - share the Takahashi/Conaway authorship line, so what looks like a converging literature is substantially one laboratory's programme. No independent group has placed MED26 at a Cajal body. The structural literature summarised in this entry's E2 is independent and does NOT resolve MED26 outside a holo-complex, so it cannot corroborate a body-level address either. Suzuki H et al. 2023
7. E_q77_loc3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review

   The proposed organising role, stated as a model rather than a measurement, is that MED26-containing Mediator recruits Cajal bodies to HISTONE LOCUS BODIES, so that 3'-end-processing factors are supplied from Cajal bodies to Pol II paused at the proximal region of transcript end sites in replication-dependent histone genes.

   MED26-containing Mediator may orchestrate multiple transcription processes through organization of nuclear bodies. · Suzuki H, Furugori K, Abe R, Ogawa S, Ito S, Akiyama T, Horiuchi K, Takahashi H · Bioessays 2023
   PMID 36852638DOI
   ▸What this evidence cannot show

   THIS IS A REVIEW and is cited as a map of the model, not as primary evidence - the primary results are [E_q77_loc2] and [E_q77_loc4]. Its own framing is hedged ('may orchestrate', 'we argue the possibility'), and this entry keeps that hedge. Authorship overlaps the primary papers.

   ▸Evidence that argues against this (1)

   • An association between two nuclear bodies, proposed in a review by the group that generated the primary data, is the weakest class of spatial claim this graph accepts. Nothing rules out that Cajal-body and histone-locus-body proximity is a consequence of shared transcriptional activity rather than of MED26 organising anything - the clustered-binding alternative the graph's condensate standards insist on. No half-bleach or model-free measurement has been applied to this system. Takahashi H et al. 2020
8. E_q77_loc4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   MED26 recruits LEC to a subset of snRNA genes through a direct interaction between EAF and the MED26 N-terminal domain; loss of MED26 decreases LEC occupancy at those genes and reduces their transcription, and the MED26 NTD functions as a molecular switch exchanging TAF7 for LEC to drive the initiation-to-elongation transition.

   MED26 regulates the transcription of snRNA genes through the recruitment of little elongation complex. · Takahashi H, Takigawa I, Watanabe M, Anwar D, Shibata M, Tomomori-Sato C, Sato S, Ranjan A, Seidel CW, Tsukiyama T, Mizushima W, Hayashi M, Ohkawa Y, Conaway JW, Conaway RC, Hatakeyama S · Nat Commun 2015
   PMID 25575120DOI
   ▸What this evidence cannot show

   Human cells; occupancy and transcription readouts rather than imaging, so this item carries the GENE-CLASS specificity and not the spatial claim - the Cajal-body element comes from [E_q77_loc2]. EAF is the partner named here; ELL is EAF's complex partner and is a KMT2A fusion partner, which is the reason this sits in a localisation block at all.

   ▸Evidence that argues against this (1)

   • The switch model is built on the SAME 87-residue N-terminal domain that docks SEC, TFIID and LEC, and this entry's own E2 records that the only atomic-detail MED26 structures are APO, isolated NMR structures of that domain with no partner in the model. So the claim that one groove sequentially binds three different complexes rests entirely on solution biochemistry, with no structure of any of the three bound states. RCSB PDB 2026
9. E_q77_loc5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   MED26's conserved N-terminal domain contains OVERLAPPING docking sites for the super elongation complex (containing ELL/EAF family members and P-TEFb), for a second ELL/EAF-containing complex, and for the general initiation factor TFIID; the proposed model is that MED26 binds TFIID in the Pol II initiation complex and then exchanges TFIID for the ELL/EAF- and P-TEFb-containing complexes to drive the transition into elongation.

   Human mediator subunit MED26 functions as a docking site for transcription elongation factors. · Takahashi H, Parmely TJ, Sato S, Tomomori-Sato C, Banks CA, Kong SE, Szutorisz H, Swanson SK, Martin-Brown S, Washburn MP, Florens L, Seidel CW, Lin C, Smith ER, Shilatifard A, Conaway RC, Conaway JW · Cell 2011
   PMID 21729782DOI
   ▸What this evidence cannot show

   Human HeLa nuclear-extract biochemistry with no spatial measurement of any kind - this is the paper that makes MED26 interesting to this project and it contains NO localisation experiment, which is worth stating explicitly in a localisation block. No fusion is present in any experiment. This item is supplied to close the gap E4 of this entry records, and E4's own wording is left unaltered because it was true of its cycle.

   ▸Evidence that argues against this (1)

   • SEC docking is a property of purified complexes in extract, and the sequential-exchange model is an inference from overlapping binding sites rather than a kinetic measurement. In the cell, the relevant question for this project is whether a KMT2A fusion's SEC is delivered through MED26 at all, and nothing in this paper or any other addresses that. The dependency data in this entry gives the reason for scepticism: KMT2A::AFF1 lines are the least MED26-dependent of the cancer groupings measured. Lee YL et al. 2021
10. E_q77_ct1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The closest published analogy for a Mediator subunit being required by a fusion oncoprotein names a DIFFERENT subunit and a DIFFERENT fusion: MED1 is required for E2A-PBX1-mediated oncogenic transcription and leukaemic cell growth. A bounded PubMed search on 2026-09-21 over MED26, CRSP7 and ARC70, and over the complex form 'Mediator complex' with MLL/KMT2A/leukaemia, returns 5 and 18 records respectively and ZERO about MED26 in any haematopoietic setting.

    Mediator subunit MED1 is required for E2A-PBX1-mediated oncogenic transcription and leukemic cell growth. · Lee YL, Ito K, Pi WC, Lin IH, Chu CS, Malik S, Cheng IH, Chen WY, Roeder RG · Proc Natl Acad Sci U S A 2021
    PMID 33542097DOI
    ▸What this evidence cannot show

    MED1 is not MED26 and E2A-PBX1 (TCF3::PBX1) is not a KMT2A fusion - this is an ANALOGY recorded to bound a rejection, not a result about MED26. The bounded search is titles and abstracts only (brief rule 80): a MED26 experiment appearing only as a supporting panel in a leukaemia paper would be invisible to it, and the complex-form query was run precisely because subunit-only enumeration of a named complex is known to under-recall (brief rule 64). The 18 complex-form hits are dominated by CDK8/19 kinase-module chemistry, which is a different module of Mediator entirely.

    ▸Evidence that argues against this (1)

    • Reading the MED1 result as support for working on MED26 would be the exact error this graph keeps finding: a conclusion imported across a subunit boundary. Mediator's kinase module, tail and head behave as separable entities, the leukaemia literature that does exist is about CDK8/19 and MED12 rather than about the middle module MED26 belongs to, and MED26's own KMT2A::AFF1 dependency is the shallowest in this entry's table. UniProt Consortium 2026

• ← related toAbsence claims: 93% are unbounded in prose, but the evidence claims are 71% bounded - and the query that tested them was itself broken

  UNBOUNDED AT ESTABLISHED: one of three absence claims carrying a confidence tier with no date, source or count. It is a structural-coverage claim about PDB rather than a literature claim, so it needs a different check than a search.

• ← related toMED12 - a named menin-associated protein that restrains menin's transcriptional output, and the anchor of the CDK8 module a KMT2A-r AML paper co-targets

  Both Mediator subunits in this graph; neither shows a KMT2A-selective requirement that survives scrutiny.

• ← related toMED23 — rank 16 of 17,728, survives the lineage split with the B-lineage half stronger, and is not absent from the leukaemia literature after all

  The graph's other Mediator entry; MED26 shows no KMT2A selectivity and MED23 ranks 16 of 17,728.

• ← supportsScoring complexes instead of genes, and the controls disagree with each other (run)

  Mediator is the strongest non-trivial complex signal, 31 subunits with 87% concordant

• ← related toThe whole protein-coding genome: menin is still first, and a metabolic pathway appears (run)

  MED13L ranks 9 and MED23 16, suggesting Mediator is under-represented in this project's interactome

• part of →Mechanism class 1 — chromatin and elongation hijack (KMT2A-r, NUP98-r)
• ← related toDiffing the core scaffold's curated interactome against the graph: 46 of 217 partners are things here, CDK9 is not one of them, and the filter I built to rank the rest was counting one platform twice

  the Mediator subunit whose neighbours have no entries

• ← regulatesMediator's KMT2A signal is in its pan-essential core, CDK8 is expressed and not required at all, and MED13L is rank 9 while its paralogue MED13 is rank 13,569

  MED26 ranks 555 and is broadly essential (pan -0.944)

• ← related toEHMT2 / G9a - the H3K9me1/2 writer that sets whether the KMT2A LOCUS ITSELF rearranges, and whose loss delays MLL-AF9 AML in a supplementary figure nobody can find

  the second Mediator subunit in the REST-to-G9a route

• ← bindsREST / NRSF - the transcription factor that delivers G9a to chromatin through MED19 and MED26, is a paediatric cancer gene, and is the worst search token in this corpus

  the second Mediator subunit in the REST-to-G9a route

• ← related toAFF4-MED10: AFF4 is prey in a 2004 Mediator MudPIT purification - a lead with an independent mechanistic reason to be true, and no experiment that tests it

  MED26 docks elongation factors and is one of the six baits in the same purification - the corroborating mechanism

• related to →ELL is in Cajal bodies - or nuclear speckles, depending which paper you read - and the MLL-ELL fusion empties them

  MED26 is the subunit that routes the ELL/EAF little elongation complex to Cajal bodies - the compartment this entry's ELL counterpart is about

• recruits →ELL — rank 7 on selectivity and the deepest common-essential in the top ten, with one NMR fragment to its name

  MED26's N-terminal TND docks ELL/EAF-containing complexes (LEC and SEC), exchanging TAF7 for them at the initiation-to-elongation transition

• related to →Transcription factories and Pol II clusters - no KMT2A::AFF1 imaging, condensate evidence for SEC and MLL-ENL, and 3D-FISH that actively contradicts the factory story for t(4;11)

  the MED26 spatial claim is about where a Pol II transition occurs, not where the protein resides - the same distinction this node draws

• related to →Nuclear speckles — the compartment everyone assumes KMT2A::AFF1 uses, on evidence that skips AFF1 itself

  the compartment MED26 has NOT been placed in: the Cajal-body result is specifically not a speckle result, and the two are routinely conflated

• ← part ofMED19 - the Mediator subunit curated against an engineered MLL-AFF4 fusion protein, and the one REST uses to deliver G9a

  the two Mediator subunits that co-purify with the MLL-AFF4 construct and the two REST uses to reach G9a

• ← related toIntAct curates 47 interactions against engineered MLL fusion proteins, and every accession-keyed query in this project is blind to all of them - while its PSICQUIC mirror stops at 2024

  MED26 carries four of the six enumerated mll_aff4 construct records - more than any other protein