Oncofusion

start / graph / import-class-human-aml-lines-for-infant-b-all

Explainermoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q-partner-provenance

Import class 4: human myeloid cell lines standing in for infant B-lymphoid disease

also called MOLM-13/THP-1 import · lineage import · AML line proxy

A lot of the human experiments in this field use adult myeloid leukaemia cell lines because they are easy to grow and easy to edit. The index disease is a B-cell leukaemia of babies, and the same fusion can produce either kind of leukaemia, so the lineage really is a variable and not a detail.

The same thing, for a scientist

This is the mildest of the five classes and the one hardest to bound, because it is a property of cell-line choice rather than of a citation. Across all 848 entries, 80 name a human myeloid line (MOLM-13, THP-1, NOMO-1, OCI-AML, ML-2) and 116 name a t(4;11) or KMT2A-r B-lymphoid line (SEM, RS4;11, MV4-11, KOPN-8, HB11;19) [E1]. So the cell-line balance is better than the fusion balance - **human index-lineage lines are used widely even where the mechanism being tested came from mouse MLL-AF9.** What makes the class real rather than pedantic is that lineage is demonstrably not downstream noise. The same conditional *Mll-AF4* allele produces B-precursor ALL in some mice and AML in others [E2]. And the downstream consequences differ: MLL-rearranged **ALL** specifically activates BCL-2 through H3K79 methylation and is sensitive to the BCL-2-selective antagonist ABT-199 [E3] - a lymphoid-restricted therapeutic consequence of the same chromatin mark. Even the CRISPR-ChIP study's human arm is **K562**, a CML line with no KMT2A rearrangement at all, used to establish H3K4me3 and Pol II biology [E4].

Go deeper
What the class does and does not threaten

Survives the lineage boundary: protein-protein contacts, domain requirements, chromatin-binding preferences, and most biochemistry. A YEATS domain reads the same acyl marks in a monocytic line and a pro-B line.

Does not survive: target gene sets, differentiation endpoints, the identity of the cell of origin, drug sensitivities mediated through lineage-restricted apoptotic machinery (the BCL-2/ABT-199 case 3), immunophenotype-based endpoints, and anything read out as 'differentiation'. Menin inhibitors are assessed clinically in part by differentiation syndrome and by monocytic maturation - endpoints with no direct lymphoid equivalent.

And an age boundary sits inside the lineage one. Infant KMT2A-r ALL arises in utero from a fetal progenitor; every human cell line in the list was derived from established disease, most from relapse, and none is fetal.

Why this class does not show up in a PMID census

The other four classes can be counted by citation, because each has defining papers. This one cannot: an entry can import a lineage without citing anything lineage-specific, simply by describing a result obtained in THP-1 as though it were a property of the disease. The 80-versus-116 string count is the best available instrument and it is weak - it counts a line named in a caveat the same as a line used for a result 1.

The honest position is that this class's size is unmeasured, and that the counting method that would measure it is per-entry reading rather than a script.

Facts and figures
classlineage and age import
entries naming myeloid lines80
entries naming kmt2a b lymphoid lines116
population848
measurablefalse
transfers well
[
  "protein-protein contacts",
  "domain requirements",
  "chromatin-reader specificity"
]
transfers badly
[
  "target gene sets",
  "differentiation endpoints",
  "lineage-restricted drug sensitivities",
  "fetal cell-of-origin biology"
]
How this could bite us back
  • Differentiation-based pharmacodynamic endpoints for menin inhibitors were developed in myeloid disease; an infant B-ALL trial that inherits them may be measuring something the lymphoid blast cannot do.
  • The graph's cell-line balance looks reassuring (116 versus 80) precisely because MV4-11, SEM and RS4;11 are easy to name in a caveat; using a line in one experiment does not make a mechanism lymphoid.
Still open
  1. How many entries in this graph state a result obtained in a myeloid line as a property of the disease without naming the line? That is the number this class needs and no script can produce it.
  2. Is there a fetal-derived KMT2A::AFF1 model at all? Every human line in the census derives from established, mostly relapsed disease, and infant leukaemia initiates in utero.
Every source behind this entry (4)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Across the 848 entries in this graph, 80 name at least one human myeloid leukaemia line from the set MOLM-13, THP-1, NOMO-1, OCI-AML or ML-2, and 116 name at least one KMT2A-rearranged B-lymphoid line from the set SEM, RS4;11, MV4-11, KOPN-8 or HB11;19.

    What this evidence cannot show

    Our own count by regular-expression match over node JSON on 2026-09-19. It counts a line named in a caveat identically to a line used for a result, and the token 'SEM' is a common abbreviation that will match incidental uses, so the B-lymphoid figure is an over-count of unknown size. MV4-11 is itself KMT2A::AFF1 but is myelomonocytic, so it sits in both categories biologically while appearing in only one here.

    Evidence that argues against this (1)
    • String counting cannot distinguish an entry that used a line from an entry that disclaimed it, and cannot see entries that describe a cell-line result without naming the line. The direction of the resulting bias is unknown, which is why this class is described as unmeasured rather than measured. Oncofusion knowledge graph et al. 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    A single conditional Mll-AF4 allele induces B-precursor acute lymphoblastic leukaemia or acute myeloid leukaemia in mice, and the resulting ALL expression profiles overlap human MLL-rearranged ALL.

    H3K79 methylation profiles define murine and human MLL-AF4 leukemias. · Krivtsov AV, Feng Z, Lemieux ME, Faber J, Vempati S, Sinha AU, Xia X, Jesneck J, Bracken AP, Silverman LB, Kutok JL, Kung AL, Armstrong SA · Cancer Cell 2008
    PMID 18977325DOI
    What this evidence cannot show

    Abstract fetched from PubMed and read on 2026-09-19; full text not read. The abstract does not report what proportion of mice develop each disease, so 'either' is qualitative. A chimeric knock-in, not the human t(4;11) product.

    Evidence that argues against this (1)
    • Bi-lineage output from one allele could equally mean lineage is a downstream choice on top of a shared fusion programme, in which case myeloid readouts transfer better than this class assumes. Yokoyama A et al. 2010
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MLL-rearranged acute lymphoblastic leukaemias activate BCL-2 through H3K79 methylation and are sensitive to the BCL-2-specific antagonist ABT-199.

    MLL-Rearranged Acute Lymphoblastic Leukemias Activate BCL-2 through H3K79 Methylation and Are Sensitive to the BCL-2-Specific Antagonist ABT-199. · Benito JM, Godfrey L, Kojima K, Hogdal L, Wunderlich M, Geng H, Marzo I, Harutyunyan KG, Golfman L, North P, Kerry J, Ballabio E, Chonghaile TN, Gonzalo O, Qiu Y, Jeremias I, Debose L, O'Brien E, Ma H, Zhou P, Jacamo R, Park E, Coombes KR, Zhang N, Thomas DA, O'Brien S, Kantarjian HM, Leverson JD, Kornblau SM, Andreeff M, Müschen M, Zweidler-McKay PA, Mulloy JC, Letai A, Milne TA, Konopleva M · Cell Rep 2015
    PMID 26711339DOI
    What this evidence cannot show

    Abstract fetched from PubMed and read on 2026-09-19; full text not read, every identifier copied from that record. 'MLL-rearranged ALL' is partner-mixed rather than t(4;11)-specific. A 36-author multicentre study, which is a strength for robustness and a weakness for attributing any single mechanistic step.

    Evidence that argues against this (1)
    • BCL-2 dependence is itself lineage-associated across leukaemias generally, so an ALL-versus-AML difference in ABT-199 sensitivity need not be caused by the fusion's chromatin programme at all. Vu LP et al. 2017
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    The human arm of the CRISPR-ChIP study is K562, annotated in GEO as source 'CML', cell type 'K562', and it supports only the H3K4me3 and Pol II conclusions; no human sample in the study carries a KMT2A rearrangement.

    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. Submitter-entered metadata. K562 is a legitimate and conventional choice for general chromatin biology; citing it here is not a criticism of the study but an illustration that 'the human experiments' in a paper need not be human experiments in the disease.

    Evidence that argues against this (1)
    • Using a non-KMT2A human line for the generic arm is good practice - it shows the COMPASS and Pol II findings are not artefacts of a leukaemic background - so this evidence item illustrates a reading hazard rather than a flaw in the source. Gilan O et al. 2023
Connected to