Oncofusion

start / graph / ana-partner-provenance-corrections

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q-partner-provenance

Corrections owed to other beats: nine entries whose stated provenance is now resolvable, and one unflagged import

also called provenance corrections cycle 49 · CRISPR-ChIP caveat fix · gene-leo1 provenance

While auditing, we found places where other entries say a fact is unknown that we were able to look up, and one place where an entry does not say which leukaemia its evidence came from. None of these is a wrong result; they are provenance statements that can now be made specific.

The same thing, for a scientist

Reported, not edited, per the standing rule. Three corrections, each with the primary source that settles it. **Correction 1 - seven entries state the CRISPR-ChIP fusion is unknown; it is murine MLL-AF9.** `hyp-menin-inhibition-spares-endogenous-h3k79-unit`, `mark-h3k79me2`, `mechanism-dot1l-af10-recruitment-axis`, `mechanism-dot1l-ppi-inhibition-modality`, `exp-h3k79-state-resolved-icechip-dose-series`, `mechanism-dot1l-catalysis-independent-function` and `mechanism-lamp5-as1-lncrna-activates-dot1l` all cite PMID 37679565, and the first five carry a caveat of the form *'the MLL translocation used is not named in the abstract'*. **It should say:** the study's deposited data identify the system as *Mus musculus*, cell type MLL-AF9, source AML - GSM5751161-2 in GSE192544 and all ten H3K79me2 ChIP-seq samples GSM5751582-91 in GSE192555 [E1][E2]. Read from GEO, not from the paper, whose full text is unreachable. **Correction 2 - the same entries understate the study's endogenous-locus evidence.** Several say the partition is *'established on an integrated reporter rather than at endogenous loci'*. The screen is reporter-based, but GSE192555 contains genome-wide H3K79me2 ChIP-seq for control, MLLT10 KO (two guides), DOT1L KO (two guides), DMSO, menin-inhibitor and DOT1L-inhibitor arms, plus FLAG-MLLAF9 occupancy ChIP-seq [E2]. **It should say:** the screen is reporter-based and the validation is genome-wide ChIP-seq in the same murine MLL-AF9 cells. **Correction 3 - `gene-leo1` (interactome2) does not record the provenance of its load-bearing citation.** Its E1 cites PMID 20541477, whose assays are MLL- and MLL-AF9-mediated activation of Hoxa9 and MLL-AF9-mediated immortalisation in a murine system; the caveat reads only *'Complex-level evidence only; LEO1 is not individually tested in this paper and the MLL contact is not mapped to a subunit.'* **It should add:** the fusion tested is MLL-AF9, not KMT2A::AFF1, and the immortalisation assay is murine - which is precisely what the sibling entry `gene-rtf1`, citing the same paper, already says [E3]. Abstract read; full text not read.

Go deeper
None of these changes a result

All three are provenance statements rather than findings. The CRISPR-ChIP partition is unchanged by knowing its system; what changes is whether an entry may extend it to KMT2A::AFF1 without a flag. LEO1's PAF1c membership is unchanged; what changes is whether a reader can tell which fusion the synergy was measured in.

The reason to report them anyway is the pattern the audit found: per-entry honesty is high and the aggregate is invisible. A caveat that says 'the source does not state its system' is honest and also terminal - it closes the question rather than queueing it. Correction 1 exists because seven entries closed the same question seven times, and the answer was a free database lookup.

A suggested rule rather than a suggested edit

Whenever an entry is about to write 'the source does not state X', where X is the experimental system, check the primary data deposition first. esearch db=gds on the paper's distinctive method name or the first author, then the GEO accession-display text endpoint. Organism and cell type are per-sample fields and are public even when the paper is not.

This would have resolved the CRISPR-ChIP question in cycle 48 rather than cycle 49, and it generalises to any paywalled paper with sequencing behind it - which in this field is most of them. Written up as concept-resolving-an-unstated-system-from-geo.

Facts and figures
scriptmanual audit, cycle 49
statusrun
result summaryThree corrections reported to other beats: (1) seven entries can now name the CRISPR-ChIP system as murine MLL-AF9; (2) the same entries understate the study's genome-wide validation arm; (3) gene-leo1 does not record that its load-bearing citation is murine MLL-AF9.
read full textfalse
notereported, not edited - every folder except partner-provenance was treated as read-only
entries affected
[
  "hyp-menin-inhibition-spares-endogenous-h3k79-unit",
  "mark-h3k79me2",
  "mechanism-dot1l-af10-recruitment-axis",
  "mechanism-dot1l-ppi-inhibition-modality",
  "exp-h3k79-state-resolved-icechip-dose-series",
  "mechanism-dot1l-catalysis-independent-function",
  "mechanism-lamp5-as1-lncrna-activates-dot1l",
  "gene-leo1"
]
How this could bite us back
  • Correcting seven caveats to say 'murine MLL-AF9' makes the graph more accurate and also makes its central hypothesis look weaker; the temptation to leave a vague caveat in place is real and should be named.
  • A correction sourced from GEO rather than from the paper inherits GEO's limits: if the submitters mis-annotated a sample, seven entries would now carry a confident wrong system instead of an honest unknown.
Still open
  1. Are there other entries in this graph whose caveat says a source 'does not state' its system, where a data deposition would settle it? A grep for that caveat phrasing would produce the work queue.
  2. Does the CRISPR-ChIP paper's figure set rest on the reporter or on the genome-wide arm for the menin result specifically? Only the full text answers that, and it is unreachable.
Every source behind this entry (3)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    GEO series GSE192544 annotates the CRISPR-ChIP H3K79me2 screen samples GSM5751161 and GSM5751162 as organism Mus musculus, source AML, cell type MLL-AF9, and its shared treatment protocol names only K562 and MLLAF9 Cas9 cells.

    CRISPR-ChIP to identify mechanisms of chromatin regulation [CRISPR-ChIP] (GSE192544) · Gilan O, Dawson MA, et al. (GEO submitters) · NCBI Gene Expression Omnibus 2023
    What this evidence cannot show

    Read on 2026-09-19 from the GEO accession-display text endpoint (targ=all&form=text&view=brief). Submitter-entered metadata, not a Methods section: it bounds the deposited sequencing rather than every experiment the authors performed. The mouse MLL-AF9 line is not further characterised in the record.

    Evidence that argues against this (1)
    • If the authors also ran non-sequencing H3K79 experiments in a human MLL-AF4 line - a western blot or a ChIP-qPCR - those would not appear in GEO, and the correction would then be understating rather than overstating the study's scope. Gilan O et al. 2023
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    GEO series GSE192555 contains genome-wide H3K79me2 ChIP-seq for control guide, MLLT10 knockout (two guides), DOT1L knockout (two guides), DMSO, menin-inhibitor-treated and DOT1L-inhibitor-treated arms, plus DMSO and menin-inhibitor-treated FLAG-MLLAF9 occupancy ChIP-seq, all annotated Mus musculus, cell type MLLAF9.

    CRISPR-ChIP to identify mechanisms of chromatin regulation [ChIP-seq] (GSE192555) · Gilan O, Dawson MA, et al. (GEO submitters) · NCBI Gene Expression Omnibus 2023
    What this evidence cannot show

    Read on 2026-09-19 from the GEO accession-display text endpoint and enumerated sample by sample (GSM5751582-5751591). Submitter-entered metadata. The menin inhibitor is not named in the record, so the correction establishes that a genome-wide menin-inhibitor arm exists without establishing the compound or exposure.

    Evidence that argues against this (1)
    • A deposited ChIP-seq sample is not proof that a given figure in the paper rests on it; the entries saying the partition is 'established on an integrated reporter' may be describing the specific figure they read about in the abstract rather than the study's full evidence. Gilan O et al. 2023
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    PMID 20541477 reports that MLL interacts directly with the PAF complex through sequences flanking the CxxC domain, that PAFc augments MLL- and MLL-AF9-mediated transcriptional activation of Hoxa9, and that MLL deletions abolishing the PAFc interaction eliminate MLL-AF9-mediated immortalisation; AFF1 and AF4 are not mentioned in the abstract.

    The PAF complex synergizes with MLL fusion proteins at HOX loci to promote leukemogenesis. · Muntean AG, Tan J, Sitwala K, Huang Y, Bronstein J, Connelly JA, Basrur V, Elenitoba-Johnson KS, Hess JL · Cancer Cell 2010
    PMID 20541477DOI
    What this evidence cannot show

    Abstract fetched from PubMed on 2026-09-19 and read in full; full text not read. The abstract does not state the species explicitly, but 'immortalisation' in this assay class and in the sibling entry gene-rtf1's reading is murine progenitor immortalisation; a full-text read would firm that up and has not been done here. **Resolved, and with a caution.** The species is confirmed **Mus musculus, female C57BL/6** - replacing an inference with a fact. But note what the series actually is: **GSE21299 is a Hoxa9-ER time course, not an MLL-fusion experiment at all**, so it bounds the species and not the fusion. 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 PAFc contact maps to the KMT2A moiety, which every fusion retains, so this is one of the claims most likely to survive the partner boundary - making the omission in gene-leo1 a provenance defect rather than a substantive error, and a low-priority correction. Yokoyama A et al. 2010
    • A deposition can answer the question asked and still be the wrong deposition for the claim. Confirming that GSE21299 is murine says nothing about which fusion the paper's leukaemia experiments used, because this series contains no fusion - it is an oestrogen-receptor-inducible Hoxa9 system. Treating a resolved species as a resolved system is the error this sweep exists to prevent. NCBI Gene Expression Omnibus 2026
Connected to