MED23 confirmed, and the zero-record claim corrected: it is not literature-free in leukaemia, an shRNA screen already found it
also called MED23 verdict · CRSP3 · KIAA1216 · DRIP130 · SUR2 · ARC130
A previous agent flagged MED23 as a promising but unfinished lead and asked for a yes or no. The answer is yes on the data — the four B-cell leukaemia lines with the KMT2A fusion need it, and the numbers are printed here line by line — but the claim that nobody has ever studied it in leukaemia is wrong. A 2016 screen in myeloid leukaemia cells found it, and that paper was missed.
The same thing, for a scientist
**Verdict: confirmed as a dependency signal, with one correction and one caveat.** Re-measured on an independent background panel, MED23 gives d = -0.480 against other leukaemia lines, decomposing to -0.314 myeloid and **-0.770 lymphoid** (z -0.734 and -1.798) — the B-lineage-stronger shape the deferring agent reported, reproduced. Pan-cancer Chronos is **-0.2096**, identical to the figure carried forward, so it is not floor-compressed. A correlation-preserving line-label permutation gives **p = 0.0024** [E1]. The four B-lineage per-line values are printed below and are not uniform: HB11;19 is an outlier at raw Chronos -2.07. **The correction: MED23 is not unstudied in leukaemia.** Searching the current symbol plus CRSP3, KIAA1216, DRIP130, SUR2 and ARC130 against KMT2A or MLL does return **zero** records, so that half of the prior claim stands [E2]. But against leukaemia generally it returns three, and one is substantive and directly on point: 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 [E3]. That paper reaches the opposite lineage from this project's signal, which makes it both corroboration and a complication.
The four B-lineage per-line values, as asked for
DepMap Chronos in the four KMT2A-rearranged B-ALL lines, raw and after the per-line global-sensitivity offset correction 1.
| line | disease | Chronos raw | line offset | offset-corrected |
|---|---|---|---|---|
| HB11;19 | B-ALL | -2.069 | -0.161 | -1.908 |
| SEM | B-ALL | -1.257 | -0.149 | -1.109 |
| RS4;11 | B-ALL | -0.880 | -0.130 | -0.750 |
| SEMK2 | B-ALL | -0.874 | -0.152 | -0.722 |
Read these before accepting the summary statistic. The mean is carried substantially by HB11;19, which is more than twice as dependent as either of the two weakest lines. Drop HB11;19 and the remaining three average -0.860 offset-corrected, which is still a real dependency but no longer a dramatic one. Three of the four are KMT2A::AFF1 or closely related B-ALL lines and SEM and SEMK2 are related to each other, so the effective independent n here is closer to three than to four.
For scale, the comparator arm's other-leukaemia mean for MED23 puts the difference at -0.770 in this half, against -0.314 in the seven myeloid lines.
Why this shape is the interesting one
This project has a named artefact class: a myeloid-enriched index arm manufacturing apparent KMT2A selectivity for genes that are really myeloid-identity dependencies. Every gene in that class is myeloid-stronger.
MED23 is B-lineage-stronger by a factor of 2.5 (-0.770 against -0.314), which is the wrong shape for that artefact and the right half for infant B-ALL, the disease this project exists for. That was the deferring agent's argument and the re-measurement supports it 1.
It is also not floor-compressed. Pan-cancer Chronos -0.2096 with 20.4% of all screened lines below -0.5 means MED23 has room to move, so the -0.770 in the B-lineage half is a genuine excursion and not a gene that everything needs being scored as selective.
The line-label permutation — gene fixed, index arm redrawn 7 myeloid + 4 lymphoid, 5,000 draws, all gene-gene correlations preserved — gives p = 0.0024. For comparison MEN1 gives 0.0002 and the ACTB offset control gives 0.092 1.
The correction, and how the previous count missed it
The claim carried forward was zero published records against MLL or KMT2A. That specific claim survives, and survives the addition of the aliases the deferring agent flagged as untested: MED23 with CRSP3, KIAA1216, DRIP130, SUR2 and ARC130, against KMT2A, MLL or 'mixed-lineage leukemia', returns zero 2.
But the broader impression that MED23 is a blank slate is wrong. Against leukaemia the same alias set returns three records, and title-checking them gives one substantive hit: Bhagwat and colleagues' 2016 study of BET inhibition and the Mediator complex in AML, in which an shRNA screen of Mediator in AML cells identified MED12, MED13, MED23 and MED24 as performing a BRD4-like regulatory function, with a shared role in sustaining a block in myeloid maturation 3. That is a functional dependency result for MED23, in acute leukaemia, in a screen.
Why it was missed: it is a complex-level paper. The title names the Mediator complex and BET bromodomains; MED23 appears only inside the abstract as one of four subunits. Any search keyed to MED23 in the title, or to MED23 with a fusion-family term, misses it. This is the same failure mode the project has already recorded for complex-name phrases being unreachable from subunit symbols — and here it cut the other way, making a studied gene look unstudied.
The other two hits are incidental: a mutual-exclusivity statistics method paper, and a C. elegans paper on the SUR-2 Mediator component where the alias matched a worm gene 2.
What the Bhagwat result does to the verdict
It cuts both ways and the entry should say so plainly.
For. An independent shRNA screen found MED23 to be functionally required in AML cells, which is external corroboration that MED23 dependency in acute leukaemia is real and not a DepMap idiosyncrasy. It also gives a mechanism with a natural connection to this graph — Mediator's role alongside BRD4 at cis-regulatory elements, sustaining a differentiation block, which is the core phenotype of KMT2A-rearranged leukaemia.
Against. The published result is myeloid, and explicitly about a block in myeloid maturation. This project's signal is B-lineage-stronger. If MED23's function in leukaemia is to sustain a myeloid differentiation block, the B-lineage excess needs a different explanation, and the most economical one is that the B-lineage excess is four cell lines, one of which is an outlier.
And it is not KMT2A-specific. The Bhagwat screen was in AML cells generally. Nothing in it says MED23 matters more with a KMT2A rearrangement than without, which is the claim this project is making.
Verdict as written for the King: MED23 is a real dependency signal with the right shape and an external functional corroboration nobody in this project had found; it is not established as KMT2A-selective, and the B-lineage excess rests on four lines. It should be promoted from weak to moderate confidence and given the Bhagwat citation, and the existing entry gene-med23 in the interactome5 folder — which this beat does not own and has not edited — should be updated by whoever does own it.
| script | analyses/fa-pathway-lineage-and-fusion-specificity/run.py |
|---|---|
| status | run |
| result summary | Confirmed. d -0.480 overall, -0.314 myeloid, -0.770 lymphoid; pan-cancer Chronos -0.2096 (not floor-compressed); line-label permutation p = 0.0024. The zero-record claim holds for KMT2A/MLL even with CRSP3, KIAA1216, DRIP130, SUR2 and ARC130 added, but is WRONG for leukaemia generally: an shRNA Mediator screen in AML already identified MED23 (PMID 27068464). |
| verdict | confirmed as a dependency signal with external functional corroboration; NOT established as KMT2A-selective; recommend promotion from weak to moderate confidence and addition of PMID 27068464 |
| uniprot | Q9ULK4 |
| length AA | 1368 |
| kmt2a records | 0 |
| leukaemia records raw | 3 |
| leukaemia records substantive | 1 |
| entry to update elsewhere | gene-med23 (data/graph/nodes/interactome5/) — not owned or edited by this beat |
inputs
[ "DepMap CRISPR Chronos via Open Targets GraphQL", "Cellosaurus KMT2A fusion annotations", "PubMed" ]
b lineage per line — the numbers behind the chart
[
{
"line": "HB11;19",
"fusion": "KMT2A-MLLT1 (MLL-ENL), per Cellosaurus",
"chronos_raw": -2.0688,
"line_offset": -0.1606,
"chronos_normalised": -1.9083
},
{
"line": "SEM",
"fusion": "KMT2A-AFF1, per Cellosaurus",
"chronos_raw": -1.2572,
"line_offset": -0.1487,
"chronos_normalised": -1.1086
},
{
"line": "RS4;11",
"fusion": "KMT2A-AFF1, per Cellosaurus",
"chronos_raw": -0.8796,
"line_offset": -0.1298,
"chronos_normalised": -0.7498
},
{
"line": "SEMK2",
"fusion": "KMT2A-AFF1, per Cellosaurus",
"chronos_raw": -0.8737,
"line_offset": -0.152,
"chronos_normalised": -0.7218
}
]aliases searched
[ "MED23", "CRSP3", "KIAA1216", "DRIP130", "SUR2", "ARC130" ]
- Mediator subunits are broadly required for transcription; a MED23 inhibitor, if one existed, would be a blunt transcriptional agent with the toxicity profile that implies.
- MED23 sits in the same functional space as BET inhibition, which has already been through the clinic in AML with limited success and substantial toxicity, so the precedent is not encouraging.
- A dependency carried by one cell line out of four is the shape that most often fails to replicate.
- Why is HB11;19 more than twice as MED23-dependent as SEMK2, and does that track fusion partner? Per Cellosaurus, HB11;19 is the only one of the four B-lineage index lines that is not KMT2A::AFF1 — it carries KMT2A::MLLT1 (MLL-ENL) — and it is also the outlier, which is either a partner effect or a coincidence of n = 4.
- Does the Bhagwat Mediator dependency reproduce in B-lineage KMT2A-rearranged cells, where this project's signal is strongest?
- Is MED23 dependency in AML KMT2A-selective, which the published screen did not ask?
- Do MED12, MED13 and MED24 — the other three subunits from that screen — show the same KMT2A-selective shape in DepMap? That is a one-hour check and it would turn a single gene into a module.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Re-measured on an independent 1,445-gene background panel, MED23 gives d = -0.480 against other leukaemia lines, -0.314 in the myeloid half and -0.770 in the lymphoid half (z -0.734 and -1.798); pan-cancer Chronos is -0.2096 with 20.4% of all lines below -0.5; and a 5,000-draw correlation-preserving line-label permutation gives p = 0.0024. The four B-lineage offset-corrected values are HB11;19 -1.908, SEM -1.109, RS4;11 -0.750, SEMK2 -0.722.
Is the Fanconi signal lineage-wide, fusion-specific, and separable from pan-essentiality? Lineage-split, subtype-control and line-label permutation re-test, with a MED23 add-on · Oncofusion knowledge graph, analyses/fa-pathway-lineage-and-fusion-specificity; DepMap Chronos via Open Targets 2026What this evidence cannot show
Four B-lineage lines, of which SEM and SEMK2 are related, so the effective n is about three. The mean is carried by HB11;19 at -1.908; without it the remaining three average -0.860. The pan-cancer figure matches the -0.2096 carried forward from the previous cycle exactly, which is reassuring for reproducibility and means it is not an independent confirmation.
Evidence that argues against this (1)
- One line contributing most of a four-line mean is the classic way a small-panel dependency looks stronger than it is. The verdict should be read as 'the signal is real in HB11;19 and present in the others', not 'B-lineage KMT2A-rearranged cells need MED23'. Ben-David U et al. 2018
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
MED23 plus the aliases CRSP3, KIAA1216, DRIP130, SUR2 and ARC130 returns zero PubMed records against KMT2A, MLL or 'mixed-lineage leukemia', and zero against acute lymphoblastic leukaemia, B-ALL or infant; against leukaemia generally it returns three, of which two are incidental on title-checking (a mutual-exclusivity statistics method, and a C. elegans paper where SUR-2 matched a worm gene).
PubMed counted-absence search for MED23 and its aliases against KMT2A/MLL and against leukaemia, run 2026-09-20 and title-checked · PubMed (NCBI) 2026What this evidence cannot show
SUR2 and ARC130 are short, cross-species aliases and inflated the count with a worm paper, which is why every hit was title-checked. A title/abstract query cannot reach a gene that appears only in a supplementary table, and this beat found exactly that failure for MED23 in E3.
Evidence that argues against this (1)
- The zero-record claim is safe only for the KMT2A conjunction. The same search strategy applied to leukaemia missed nothing but very nearly buried the one paper that matters, which appears in an abstract naming the Mediator complex rather than MED23. Bhagwat AS et al. 2016
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
An shRNA screen of the Mediator complex in AML cells identified MED12, MED13, MED23 and MED24 as performing a regulatory function similar to BRD4, including a shared role in sustaining a block in myeloid maturation — so MED23 is an already-published functional dependency in acute leukaemia.
BET Bromodomain Inhibition Releases the Mediator Complex from Select cis-Regulatory Elements. · Bhagwat AS, Roe JS, et al. · Cell Rep 2016What this evidence cannot show
Read from the abstract, not the full text, in this session. The screen is in AML cells and the phenotype described is a myeloid maturation block; KMT2A rearrangement status of the cells is not stated in the abstract, and no comparison against non-KMT2A leukaemia is made. **Resolved, and it confirms the caveat's second half as a real absence.** The dataset is **48 of 48 murine MLL-AF9/NrasG12D (RN2)** - so the KMT2A rearrangement status is answered, yes - **and there are zero non-KMT2A comparator samples in it**. A B-lineage verdict is therefore being drawn from a single-genotype myeloid dataset with no comparator arm. GSE74651 and four related series. Resolved 2026-09-19 by the Q18b unstated-systems sweep from the study's deposition, not from its text; bounded as a statement about the deposited data rather than about the whole study.
Evidence that argues against this (2)
- The lineage is wrong for this project's signal. If MED23's leukaemia function is to sustain a myeloid differentiation block, it does not explain a B-lineage-stronger dependency, and the B-lineage excess may simply be four cell lines with one outlier. Oncofusion knowledge graph, analyses/fa-pathway-lineage-and-fusion-specificity 2026
- Selectivity cannot be estimated from a dataset with one genotype. A gene scoring as required in 48 samples that are all MLL-AF9/NrasG12D is required in that context, and the inference to B-lineage KMT2A::AFF1 disease crosses both lineage and fusion with no internal control to bound it. NCBI Gene Expression Omnibus 2026
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
MED23 is UniProt Q9ULK4, 1,368 aa, primary gene name MED23, Mediator of RNA polymerase II transcription subunit 23, with synonyms ARC130, CRSP3, DRIP130, KIAA1216 and SUR2.
UniProtKB entry Q9ULK4 (MED23) · UniProtKB/Swiss-Prot 2026What this evidence cannot show
Resolved by a gene_exact query and read back against the primary gene name and length, because gene_exact matches synonyms. The alias list here is the source of the five symbols searched in E2.
Evidence that argues against this (1)
- UniProt's synonym list is not exhaustive for historical literature, and the 1990s Mediator literature used complex-specific nomenclature (CRSP, DRIP, ARC, TRAP) in which individual subunits are named only by apparent molecular weight, so further older work may be unreachable by any symbol query. UniProtKB/Swiss-Prot 2026
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)
- related to →KMT2A / MLL1 — the gene that breaks
zero PubMed records for MED23 plus CRSP3, KIAA1216, DRIP130, SUR2 and ARC130 against KMT2A or MLL — the prior zero-record claim survives the alias expansion
- depends on →KMT2A-rearranged acute lymphoblastic leukaemia
the four B-lineage per-line values: HB11;19 -1.908, SEM -1.109, RS4;11 -0.750, SEMK2 -0.722 (offset-corrected)
MED23 was measured as an add-on to the same run
the correction: an shRNA Mediator screen in AML found MED12, MED13, MED23 and MED24 to act like BRD4 at cis-regulatory elements, sustaining a myeloid maturation block (PMID 27068464)
the systematic subunit scan behind the single-subunit verdict reached earlier
resolves the cells as KMT2A-rearranged and establishes that no non-KMT2A comparator was deposited