The mouse fusion fraction is 0.73 and the human proxy fusion fraction is 0.21-0.24 - a three-fold gap in the direction a retroviral over-expression model predicts, so the one place the masking factor can be calibrated against a tag is the one place its value does not transfer
also called the calibration does not transfer · 0.73 against 0.21
The mouse model that carries a clean tag also carries far more of the cancer protein than a human patient's cells do, because it was put there artificially. So the number you can measure cleanly is not the number you need.
The same thing, for a scientist
The dispatch brief for this beat reasoned: *GSE260742 has a FLAG tag, so it is the only place the masking factor can be calibrated against ground truth; if FLAG and MLL1 diverge in mouse, that quantifies how much of the human MLL1-N signal is wild-type.* **The first half is right and the second does not follow**, and this beat can now say by how much [E1]. | system | fusion's share of the MLL1-N channel at the fusion's own sites | |---|---:| | murine FLAG-MLLAF9, against a FLAG tag | **0.7317** [0.7025, 0.7586] | | human RS4;11 (KMT2A::AFF1), menin as the fusion proxy | **0.212** | | human MOLM13 (KMT2A::MLLT3), menin as the fusion proxy | **0.2373** | **Three-fold apart, and every term in the comparison is contestable.** The mouse figure comes from a model whose *own control refutes the model that produced it*; the human figures substitute menin for a tag, and menin is itself an antibody readout that cannot distinguish menin on the fusion from menin on wild-type KMT2A/KMT2B. **But the direction is not contestable and it is mechanistic.** The murine cells are c-KIT+ bone marrow retrovirally transduced with a Flag-MLLAF9 construct, so fusion protein is over-expressed relative to endogenous wild-type Mll1. The human lines carry a single rearranged allele against a wild-type allele. **A higher fusion-to-wild-type ratio makes the MLL1-N channel more fusion-dominated and so more responsive to the drug**, which is exactly the observed 0.40 in mouse against 0.94 in human MOLM13 for the same antibody. **So the transferable content of the mouse arm is qualitative and it is still valuable:** *the shared-epitope antibody does diverge from the tagged fusion, at two independent loci and at genome scale, in the direction the masking hypothesis requires.* The **magnitude** is a property of the model system and the human magnitude remains unmeasured.
What would fix it, in order of cost
- An MLL1 C-terminal ChIP in any human KMT2A-rearranged line, beside the existing N-terminal track. The C-terminus is lost in the fusion, so the C-terminal track is wild-type-only and the difference is the fusion. This is one extra ChIP in an existing experiment and no dataset in this project's evidence base has done it.
- An endogenous knock-in tag at the fusion allele in a human line - this graph already holds
concept-epitope-tagged-knock-in-at-the-endogenous-fusion-alleleandexp-q118-a-tagged-kmt2a-aff1-chip-in-rs4-11-under-vtp-50469. Correct, expensive, and the only route that gives the number directly. - A junction-specific antibody. This project has costed one (clone TEL 6F2) and not raised it.
Route 1 is an order of magnitude cheaper than route 2 and would bound the answer, and the reason nobody runs it is that the field converged on one N-terminal reagent - the same Bethyl A300-086A - in both of the datasets this project depends on.
| script | analyses/q120-h3k79me2-second-line-and-the-mll1-epitope/addendum-rank-and-crossline.py |
|---|---|
| status | run |
| result summary | Mouse fusion fraction 0.7317, human menin-proxy fusion fraction 0.212-0.2373; same antibody, MLL1 held-set ratio 0.3303 in mouse against 0.9438 in the human line carrying the same fusion. |
| antibody | Bethyl A300-086A in both systems |
| shown for fusion | MLL-AF9 in mouse (tagged, over-expressed) against KMT2A::AFF1 and KMT2A::MLLT3 in human (endogenous, untagged) |
| absence class | never-measured |
| absence class note | The human fusion fraction of the MLL1-N ChIP channel has never been measured. No archive can supply it: it needs either a C-terminal MLL1 ChIP or a tagged human allele. The murine calibration exists and this entry measures why it cannot substitute. |
| absence class assigned | 2026-09-23 (Q120) |
- Every mouse model of a fusion oncoprotein built by over-expression inflates the fusion's share of any shared-epitope readout, so the masking problem is systematically UNDER-estimated by the models built to study it and over-estimated nowhere.
- What is the actual fusion-to-wild-type protein ratio in the murine FLAG-MLLAF9 line? A single western with an N-terminal MLL1 antibody would give it and would convert this qualitative argument into a quantitative correction.
- Do any published human KMT2A-rearranged data use an MLL1 C-terminal antibody for ChIP? A bounded literature and archive search for that reagent class was not run in this beat and is the cheapest next step for the whole masking question.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
The naive fusion fraction of the MLL1-N channel is 0.7317 [0.7025, 0.7586] in murine FLAG-MLLAF9 measured against a FLAG tag, and 0.212 in human RS4;11 and 0.2373 in human MOLM13 when the same formula is applied with menin as the fusion proxy; the same antibody gives an MLL1 held-set share ratio of 0.3303 in mouse and 0.9438 in human MOLM13, which carries the same fusion.
Q120 re-analysis of GSE260742: GSM8123395/97 binned at 2 kb and compared with Q117's binned FLAG tracks in the same universe; analyses/q120-h3k79me2-second-line-and-the-mll1-epitope/arm2-mouse-epitope.py and results-arm2.json, run 2026-09-23 · Iqbal Lab oncofusion knowledge graph 2026What this evidence cannot show
The human figures use menin as a proxy for the fusion, which is a proxy and not a calibration: Bethyl A300-105A anti-Menin cannot distinguish menin bound to the fusion from menin bound to wild-type KMT2A/KMT2B, nor from menin's non-KMT2A functions. The mouse figure comes from a two-component model that this beat's own control set refutes, so it is at best an upper bound on the fusion fraction in that system. Comparing a mouse retroviral over-expression model with human endogenous fusions also crosses species, lineage, drug dose (500 nM against 330 nM) and exposure. n = 1 per antibody per condition per line in both series. No number in this beat carries a spread and none is precise.
Evidence that argues against this (1)
takes the mouse calibration and measures why its magnitude does not transfer to human
the retroviral over-expression alternative inflates the fusion's share of any shared-epitope readout three-fold, which is the argument for an endogenous knock-in
the same antibody gives 0.3303 at the held set in mouse and 0.9438 in the human line carrying the same fusion - the inertness is a property of the human system's fusion-to-wild-type ratio, not of the reagent alone
the human-side measurement whose absence is why the mouse calibration cannot be transferred