MED23 — 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
also called MED23 · SUR2 · CRSP3 · DRIP130 · ARC130 · KIAA1216 · Mediator subunit 23
MED23 is a Mediator subunit that, in this project's dependency data, looks like one of the most KMT2A-selective genes in the entire genome - and unlike most such hits it holds up when the leukaemia cell lines are split into myeloid and B-lineage groups. Not one published paper mentions it together with KMT2A.
The same thing, for a scientist
MED23 (Q9ULK4, 1,368 aa) is a Mediator tail-module subunit with no UniProt-annotated domain boundaries across 1,368 residues despite 13 PDB cross-references [E3]. It is in this beat because of a data result, stated plainly as such. In the project's genome-wide scan it ranks **16 of 17,728** on KMT2A selectivity, difference **-0.4827** (KMT2A-rearranged mean -0.9829 against other-leukaemia mean -0.5002) on a pan-essentiality of only **-0.2096**, with an empirical p against the scan distribution of **0.001185** [E1]. More unusually, it **survives the lineage split** - the only gene in the 23-gene panel besides MEN1 and KMT2A itself to do so - with **myeloid z -0.77 and B-lineage z -1.85**, the B-lineage half being the *stronger* one [E2]. That inversion matters: this beat's standing warning is that a myeloid-enriched index panel manufactures myeloid-identity hits for free, and MED23 is the opposite shape. Against all of that, **`(MLL OR KMT2A) AND (MED23 OR SUR2 OR CRSP133 OR DRIP130 OR ARC130)` returns 0 records**, and `(MED23 OR SUR2) AND (leukemia OR leukaemia OR AML)` returns 4, none of which is a MED23 leukaemia study on title-check [E4]. This entry is therefore a **data-driven lead with no literature**, and it says so in every field.
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.
I have never read one of these pictures before
A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.
Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.
Anything in yellow sticks is a bound small molecule, which is usually the drug.
The number, and every reason to distrust it
The case for looking at MED23 is unusually good by this graph's standards, and the case against believing it is also good. Both are set out here because the entry's value lies in the tension.
For:
- Rank 16 of 17,728, difference -0.4827, empirical p 0.001185 against the scan's own distribution 1.
- Pan-essentiality of only -0.2096, so the effect is not floor-compressed - the comparator arm is barely dependent and the index arm is clearly dependent. This is the shape a selective dependency is supposed to have, and almost nothing else in this beat has it.
- It survives the lineage split, with the B-lineage arm stronger (z -1.85) than the myeloid arm (z -0.77) 2. For an infant-B-ALL project that is the right half to be strong in.
Against:
- Zero published records tie MED23 to KMT2A or MLL 4. A genome-wide top-20 hit in a well-studied disease with no literature at all is more often an artefact than a discovery.
- Four B-lineage index lines produce the z of -1.85. One line could largely generate it.
- The entire Mediator panel ranks improbably well on this statistic - MED23 16, MED1 81, MED24 128, MED12 131, MED14 453 1 - and MED1 is a demonstrated myeloid-identity artefact 2. When a whole complex ranks high and one member is a proven artefact, the prior that the rest share a common cause rises sharply.
- No permutation test, lineage-conditioned analysis or line-normalised analysis was run on MED23 in this cycle, although scripts for all three exist in this repository.
The honest position: MED23 is the single best-shaped unexplained number this beat produced, and it has not been tested enough to be believed. It is recorded at weak overall confidence with the counts and caveats attached, precisely so that a future cycle can either confirm it or kill it cheaply.
What would settle it, in order of cost
- Read the per-line values. The z of -1.85 comes from four lines. Printing them individually would immediately show whether one line carries the signal. This costs minutes and was not done in this cycle - the largest shortcut this entry takes.
- Run the three disciplining analyses already in the repository - permutation, lineage conditioning, line normalisation. Round 4 was criticised in its own enumeration ledger for skipping these; this entry repeats that criticism against itself.
- Check expression. If MED23 is simply more expressed in KMT2A-rearranged lines, the dependency may follow expression rather than fusion biology.
- A single CRISPR validation in SEM and RS4;11 against a non-KMT2A B-ALL line. That is the experiment that converts this from a number into a result, and it is small.
Only step 4 requires bench work. Steps 1 to 3 are analyses this project can run in an afternoon, which is why leaving this entry at weak rather than pursuing it is recorded below as the biggest deliberate gap in this beat.
Cell type, cell state, localisation, epigenetics - all absent, stated explicitly
The standing saturation priorities require these axes to be filled where evidence exists and stated as absent where it does not. For MED23 in this disease, every one is absent:
- Cell type: none. The only cells in which MED23 has been observed in a KMT2A context are DepMap cell lines, in a dependency screen, with no biology attached.
- Cell state: not established.
- Subcellular: nucleus, by UniProt annotation 3. That is the only localisation fact available.
- Subnuclear: not established.
- Epigenetic readout: none. MED23 has no catalytic activity and no chromatin readout has been measured for it here.
- Domain structure: UniProt annotates no domains, motifs, zinc fingers, coiled coils or repeats with residue coordinates across 1,368 residues, despite 13 PDB cross-references 3. This entry therefore contributes no domain lane at all, rather than estimating boundaries from a structure.
An entry this empty would normally not be written. It is written because the dependency number is the best-shaped one in the beat and because an unexplained top-20 hit that nobody records is an unexplained top-20 hit that gets rediscovered and re-forgotten every cycle.
| uniprot | Q9ULK4 |
|---|---|
| gene symbol | MED23 |
| protein name | Mediator of RNA polymerase II transcription subunit 23 |
| length AA | 1368 |
| alphafold ID | AF-Q9ULK4-F1 |
| accession resolution | UniProt REST query gene_exact:MED23 AND organism_id:9606 AND reviewed:true, run 2026-09-19; one reviewed hit; primary gene name read back as MED23 and length 1368 confirmed from the same fetched record. |
| alias hazard | SUR2 collides with a C. elegans gene and inflates the count. The enumeration searched CRSP133 rather than the UniProt-listed CRSP3, so the zero-record claim has a known hole in its symbol set. |
| complex | Mediator, tail module |
| domains note | UniProt annotates NO features with residue coordinates across 1,368 residues despite 13 PDB cross-references. No domain lane is contributed rather than estimating boundaries from a structure. |
| function | Mediator tail-module subunit; no catalytic activity |
| classification | a data-driven lead with no literature; NOT a documented interactor of KMT2A::AFF1 or of anything else in this graph |
| interaction evidence class | none - the entry rests entirely on a CRISPR dependency screen |
| shown for fusion | none |
| cell type evidence | none beyond DepMap cell lines in a dependency screen |
| cell state evidence | not established |
| epigenetic readout | none |
PDB IDS
[ "6H02", "7EMF", "7ENA", "7ENC", "7ENJ", "7LBM", "8GXQ", "8GXS", "8T9D", "8TQW", "8TRH", "9F6Y", "9F76" ]
aliases retired or alternate
[ "SUR2", "CRSP3", "DRIP130", "ARC130", "KIAA1216" ]
complex partners
[ "gene-med12", "gene-med26" ]
domains
[]
localization
{
"subcellular": "nucleus (UniProt)",
"subnuclear": "not established"
}essentiality
{
"depmap_scan": "analyses/depmap-kmt2a-selective-scan-full",
"pan_essentiality_all_lines": -0.2096,
"n_all_lines": 1208,
"mean_kmt2a_rearranged": -0.9829,
"mean_other_leukemia": -0.5002,
"kmt2a_selectivity_difference": -0.4827,
"rank": 16,
"n_scanned": 17728,
"percentile": 0.09,
"empirical_p_vs_scan": 0.001185,
"lineage_split": {
"analysis": "analyses/depmap-interactome5-lineage-split",
"z_kmt2a_myeloid": -0.77,
"z_kmt2a_b_lineage": -1.85,
"z_split_gap": 1.08,
"z_OCI_AML3": -1.58,
"z_U937": 0.14,
"n_kmt2a_b_lineage_lines": 4,
"verdict": "lineage-independent - one of only three survivors in the panel, with MEN1 and KMT2A"
},
"not_floor_compressed": true,
"analyses_not_run": [
"permutation test",
"lineage-conditioned analysis",
"line-normalised analysis",
"per-line value inspection",
"expression normalisation"
],
"interpretation": "the best-shaped unexplained dependency in this beat and the least tested; recorded at weak confidence so a future cycle can confirm or kill it cheaply"
}enumeration record counts
[
{
"query": "(MLL OR KMT2A) AND (MED23 OR SUR2 OR CRSP133 OR DRIP130 OR ARC130)",
"records": 0,
"note": "symbol set omits CRSP3 and KIAA1216 - a known hole"
},
{
"query": "(MED23 OR SUR2) AND (leukemia OR leukaemia OR AML)",
"records": 4,
"records_after_title_check": 0
},
{
"date": "2026-09-19"
}
]depmap kmt2a selectivity
{
"chronos_mean_kmt2a_rearranged": -0.9829,
"chronos_mean_other_leukaemia": -0.5002,
"difference": -0.4827,
"pan_essentiality_all_lines": -0.2096,
"selectivity_rank": 16,
"rank_denominator": 17728,
"provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
"how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}- MED23 is a subunit of general transcription machinery. Even if the dependency is real, the therapeutic window against normal proliferating tissue is unknown and the pan-essentiality of -0.21 measures tolerance in cell lines, not in a child.
- There is no structure-derived handle: UniProt gives no domain coordinates at all, so a structure-based programme would be starting from nothing despite the protein appearing in solved Mediator assemblies.
- A top-20 genome-wide hit with zero supporting literature is exactly the profile of an artefact, and the Mediator panel it belongs to contains a demonstrated lineage artefact. Acting on it before the cheap analyses are run would be the classic error this graph was built to prevent.
- The zero-record enumeration behind this entry has an acknowledged hole in its symbol set (CRSP3 and KIAA1216 unsearched), so even the claim that nothing is published is provisional.
- Is MED23 more expressed in KMT2A-rearranged lines? If the dependency tracks expression it is probably not fusion biology.
- Why does a gene ranking 16 of 17,728 in a heavily studied disease have zero published records against MLL or KMT2A? Either the field has missed it or the number is an artefact, and those are very different worlds.
- Does MED23 knockout affect KMT2A::AFF1 target gene expression, or only viability? Viability without a transcriptional phenotype would point to a general-proliferation confound.
- HB11;19 carries the lymphoid signal and is KMT2A::MLLT1 rather than KMT2A::AFF1. Recompute with it excluded before MED23 is described as a B-lineage dependency.
- MED12 and MED24 were identified in the same shRNA screen and this graph holds a MED12 entry. Compare their lineage-split profiles before treating MED23 as subunit-specific.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database
In the project's genome-wide KMT2A-rearranged selectivity scan of 17,728 genes, MED23 ranks 16 (0.09th percentile) with a difference of -0.4827 (KMT2A-rearranged mean -0.9829 against other-leukaemia mean -0.5002) on a pan-essentiality of -0.2096 across 1,208 lines, with an empirical p against the scan distribution of 0.001185. Other Mediator subunits in the same scan: MED1 -0.3114 rank 81, MED24 -0.2698 rank 128, MED12 -0.268 rank 131, MED14 -0.1541 rank 453.
Genome-wide KMT2A-rearranged selectivity scan, 17,728 genes (DepMap Chronos via Open Targets) · Broad Institute DepMap / Open Targets Platform; analyses/depmap-kmt2a-selective-scan-full 2026What this evidence cannot show
Rows read out of analyses/depmap-kmt2a-selective-scan-full/results.json on 2026-09-19 in a scripted pass; no number typed from recall. Eleven KMT2A-rearranged against 43 other leukaemia lines, AML-heavy, none infant-derived. **The empirical p is against the scan's own distribution, which is not a permutation test**, and no permutation, lineage-conditioned or line-normalised analysis was run on MED23 in this cycle despite scripts existing for all three. The low pan-essentiality means this is NOT a floor-compressed number, which is the strongest thing about it.
Evidence that argues against this (1)
- The entire Mediator panel ranks inside the top 460 of 17,728, and MED1 - rank 81 - is demonstrated by the lineage split to be a myeloid-identity artefact. When one member of a co-ranking complex is a proven artefact, the prior that the others share a common cause such as transcriptional output or proliferation rate rises substantially. Broad Institute DepMap / Open Targets Platform; analyses/depmap-interactome5-lineage-split 2026
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database
On the lineage-split run for this beat, MED23 is one of only three genes in the 23-gene panel classified 'lineage-independent', alongside MEN1 and KMT2A: KMT2A-rearranged myeloid z -0.77, B-lineage z -1.85, split gap +1.08, OCI-AML3 z -1.58, U-937 z +0.14, pan-essentiality -0.206. MEN1 in the same run is myeloid z -2.29 and B-lineage z -2.28; KMT2A is -1.48 and -1.22. MED1 by contrast is flagged a MYELOID-IDENTITY ARTEFACT at myeloid z -0.86 against B-lineage z -0.13.
Interactome round 5 panel, lineage-split DepMap dependency (FACT, NSL, MSL, COMPASS remainder, Mediator) · Broad Institute DepMap / Open Targets Platform; analyses/depmap-interactome5-lineage-split 2026What this evidence cannot show
Rows read from analyses/depmap-interactome5-lineage-split/results.json on 2026-09-19. **Four B-lineage index lines generate the z of -1.85, so a single line could largely produce it, and the per-line values were not printed in this cycle.** The split is by lineage LABEL rather than expression signature. Raw means are quoted where available; the z values are in other-leukaemia SDs and are not effect-size comparisons across genes.
Evidence that argues against this (1)
- MEN1 achieves near-identical z in both halves (-2.29 and -2.28), which is what a genuinely lineage-independent dependency looks like. MED23's halves differ by 1.08, so calling it lineage-independent is a threshold decision rather than an observation of symmetry, and the label should not be over-read. Broad Institute DepMap / Open Targets Platform; analyses/depmap-kmt2a-selective-scan-full 2026
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
UniProtKB Q9ULK4 (MED23_HUMAN, Mediator of RNA polymerase II transcription subunit 23) is 1,368 aa with synonyms ARC130, CRSP3, DRIP130, KIAA1216 and SUR2; it is annotated to the nucleus; it has 13 PDB cross-references; and it carries no annotated domains, zinc fingers, repeats, coiled coils, motifs or DNA-binding regions with residue coordinates.
UniProtKB Q9ULK4 (MED23_HUMAN), Mediator of RNA polymerase II transcription subunit 23 · UniProt Knowledgebase, reviewed (Swiss-Prot) 2026What this evidence cannot show
Record fetched 2026-09-19 via UniProt REST with gene_exact:MED23 AND organism_id:9606 AND reviewed:true; **primary gene name read back as MED23 and length 1,368 confirmed from the same fetched record**, per the standing accession rule. One reviewed hit. An absence of annotated features reflects curation state, not absence of structure - 13 PDB entries exist, most of them Mediator assemblies in which MED23 is one chain.
Evidence that argues against this (1)
- Because no feature has coordinates, this entry contributes no domain lane and no structural reasoning is possible. Any downstream ligandability assessment that treated MED23 as structurally characterised would be inventing content, even though the protein does appear in solved Mediator structures. Broad Institute DepMap / Open Targets Platform; analyses/depmap-kmt2a-selective-scan-full 2026
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database
As of 2026-09-19, PubMed returns 0 records for '(MLL OR KMT2A) AND (MED23 OR SUR2 OR CRSP133 OR DRIP130 OR ARC130)' and 4 for '(MED23 OR SUR2) AND (leukemia OR leukaemia OR AML)'. Title-checking the 4 leaves none that is a MED23 leukaemia study: a zebrafish survivin2 primitive-haematopoiesis paper, a BET-bromodomain/Mediator study, a C. elegans EOR-1/PLZF Ras and Wnt signalling paper matched on the SUR-2 worm gene, and a mutual-exclusivity statistical methods paper. **This absence claim is now bounded and partly withdrawn.** The queries are symbol-driven, and MED23's leukaemia literature is COMPLEX-LEVEL: an shRNA screen of Mediator in AML cells identified MED12, MED13, MED23 and MED24 as performing a BRD4-like regulatory function, including a shared role in sustaining a block in myeloid maturation (PMID 27068464). That paper is unreachable from any MED23 symbol query because its abstract frames the screen at the level of the complex — the same failure class as the NUP98/NSL paper being unreachable from KANSL symbols. The KMT2A-specific absence survives; the leukaemia absence does not.
PubMed E-utilities esearch record counts for the Mediator subunits MED1, MED12, MED14, MED23 and MED24 against KMT2A, MLL and leukaemia, run for interactome round 5 · NCBI PubMed, queried 2026-09-19 2026What this evidence cannot show
Counts from the esearch `count` field, queried 2026-09-19, with **every record in the 4-record query title-checked individually**. Symbol set: MED23, SUR2, CRSP133, DRIP130, ARC130. Note that CRSP3 - the UniProt-listed synonym - was NOT in the searched set; CRSP133 was searched instead, which is a spelling this beat carried in from the alias list and is a genuine gap. The alias SUR2 collides with a C. elegans gene and inflates rather than improving recall.
Evidence that argues against this (2)
- The searched alias set omitted CRSP3 and KIAA1216, so this zero is a floor with a known hole in it. A zero-record claim is only as good as its symbol set, and this one is explicitly imperfect - which is the honest way to report it and a reason to re-run the enumeration before acting on the dependency. UniProt Knowledgebase, reviewed (Swiss-Prot) 2026
- A symbol-driven absence claim cannot see a complex-level abstract, and this one did not. The correct statement is narrower: no study addresses MED23 in KMT2A-rearranged leukaemia specifically, while MED23 in AML has been screened as part of Mediator. Bhagwat AS et al. 2016
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
An shRNA screen of the Mediator complex in acute myeloid leukaemia cells identified MED12, MED13, MED23 and MED24 as performing a similar regulatory function to BRD4, including a shared role in sustaining a block in myeloid maturation, in a study showing that BET inhibition rapidly releases Mediator from a subset of cis-regulatory elements enriched for MYB targets and leukaemogenesis functions.
BET Bromodomain Inhibition Releases the Mediator Complex from Select cis-Regulatory Elements · Bhagwat AS, Roe JS, Mok BYL, Hohmann AF, Shi J, Vakoc CR · Cell Reports 2016What this evidence cannot show
Abstract read; full text not read. The screen is in AML cells and the paper does not stratify by KMT2A rearrangement, so it establishes MED23 as a Mediator subunit with a leukaemia phenotype and not as a KMT2A-selective dependency.
Evidence that argues against this (2)
- The phenotype is a MYELOID maturation block, and this graph's own lineage split puts MED23's KMT2A selectivity in the B-LYMPHOID half (−0.770 lymphoid against −0.314 myeloid). If the published mechanism is myeloid maturation, it is probably not the mechanism behind the dependency this beat found. Oncofusion knowledge graph (internal analysis) 2026
- Four Mediator subunits scoring alike in one shRNA screen is the profile of a complex-level requirement rather than a subunit-specific one, and this graph separately records that a subunit shared across complexes cannot be inhibited without affecting all of them. Bhagwat AS et al. 2016
- E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.analysis
On a line-permutation test that holds the gene fixed and redraws the index arm as 7 myeloid and 4 lymphoid lines, MED23 gives d −0.480 at p = 0.0024, with the B-lineage half stronger (−0.770) than the myeloid half (−0.314) and pan-cancer Chronos −0.2096. The four offset-corrected B-lineage values are HB11;19 −1.908, SEM −1.109, RS4;11 −0.750 and SEMK2 −0.722.
Fanconi pathway lineage split and fusion specificity in KMT2A-rearranged leukaemia lines · Oncofusion knowledge graph (internal analysis) 2026What this evidence cannot show
Four lines carry the lymphoid half. HB11;19 is, per Cellosaurus, the only one of the four that is NOT KMT2A::AFF1 — it carries KMT2A::MLLT1 — and it also carries the largest single value.
Evidence that argues against this (2)
- The mean is carried by one line. HB11;19 at −1.908 is nearly twice the next value, and removing it would move the lymphoid half substantially, so the 'B-lineage stronger' reading rests on a line that is not the index fusion. Oncofusion knowledge graph (internal analysis) 2026
- The open question this entry recorded last cycle — do the four B-lineage lines agree, or does one carry the z — is now answered, and the answer is the unfavourable one. The rank survives; which lineage it belongs to does not. Oncofusion knowledge graph (internal analysis) 2026
COUNTED ABSENCE RE-RUN AND HOLDS: 69 free-text hits, none in title or abstract with KMT2A.
verdict: confirmed as a dependency signal (d -0.480; -0.770 B-lineage, -0.314 myeloid; line-permutation p = 0.0024; pan-cancer Chronos -0.2096, not floor-compressed)
Same complex, opposite behaviour on the lineage split: MED12 is myeloid-weighted, MED23 survives in both halves.
Same complex; MED23 survives the lineage split while MED12 is myeloid-weighted and MED1 is a demonstrated artefact.
The graph's other Mediator entry; MED26 shows no KMT2A selectivity and MED23 ranks 16 of 17,728.
One of only three genes in this beat's 23-gene panel to survive the lineage split, alongside MEN1 and KMT2A.
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
Data-driven lead only: rank 16 of 17,728 with a stronger B-lineage than myeloid arm, and zero published records tying MED23 to KMT2A.
The only survivor of the split with no literature, and the beat's largest untested lead.
MED23 is the best-ranked core subunit at 16, and is broadly essential
MED23 was the top unique ZNF384-fusion interactor confirmed by co-immunoprecipitation; wild-type ZNF384 did not interact