Oncofusion

start / graph / ana-the-index-fusion-line-holds-2-5x-more-h3k79me2-under-one-dot1l-inhibitor-and-less-under-the-other

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q138-absolute-h3k79me2

RS4;11 retains 2.5x more H3K79me2 than MOLM13 under EPZ5676 at every one of ten baseline deciles - and 0.71x under compound 10, with the sign reversing in deciles 5-10: the between-line difference survives the baseline confound and fails the drug replication, so it cannot be attributed to the fusion partner

also called the partner difference does not replicate · baseline-robust, drug-labile · C1, C2 and C3 all fail

Comparing the two leukaemia lines, one appears to hold on to nearly three times more of the chromatin mark under one drug - a difference that survives every check for the obvious artefact. Then the second drug reverses it. A difference that flips sign when you change the drug is not a property of the cells' fusion gene.

The same thing, for a scientist

**Statement, ABSOLUTE for the levels and RELATIVE for the ranks.** On the 16,636 TSS entries measurable in both lines at day 9 (of 19,214 shared; 2,578 dropped for a baseline below 25 in either line, 13.4%, and that count is printed): | contrast | RS4;11 median retained | MOLM13 median retained | ratio | |---|---:|---:|---:| | EPZ5676 1 uM, d9 | **3.93%** | **1.60%** | **2.53x** | | compound 10 100 nM, d9 | **1.70%** | **2.46%** | **0.71x** | **The EPZ difference is not a baseline artefact.** Conditioned on matched MOLM13 baseline decile it holds in **all ten deciles**, with a median log2 difference from **+0.82 (decile 10) to +2.25 (decile 4)** - never crossing zero. That is exactly the check pre-registered branch C3 demanded, and the difference survived it. **And then compound 10 reverses it.** Decile 1 -0.15, deciles 2-4 +0.24 to +0.59, deciles 5-10 **-0.08 to -0.97**. The two drugs do not agree on which line holds more, and they disagree within the same deciles. **Per-gene shape agrees almost perfectly.** The Spearman correlation of per-gene log2 absolute change between the lines is **0.785** (EPZ) and **0.751** (compound 10), against a ceiling set by the similarity of their baselines, **0.815**. Once a gene's starting level is known, the two lines lose H3K79me2 in nearly the same order. **None of C1, C2 or C3 describes this**, and it is logged as a pre-registration failure rather than rewritten. C1 wanted a systematic difference; there is one and it inverts. C2 wanted agreement beyond the global factor; the ranks agree and the levels do not. C3 wanted a baseline confound; the difference survives the baseline conditioning and dies on drug replication, which no branch named.

Go deeper
The bound that was declared before the numbers were seen

RS4;11 and MOLM13 differ in fusion partner (AFF1 vs MLLT3), lineage (B-ALL vs acute monocytic leukaemia), donor, sex (female vs male) and age (32Y vs 20Y), and RS4;11 additionally carries an NSD2 p.Glu1099Lys chromatin lesion. A difference between these two lines is not attributable to the partner, and the pre-registration said so before any number was computed.

The drug reversal makes the point moot in the most useful way: the difference is not even attributable to the lines, because the same two lines swap places when the compound changes. Whatever produces it is a drug-by-line interaction, and with one library per cell it cannot be separated from a library-level effect. 12

What this says about importing MLLT3 biology into the index fusion

This project has imported nearly all its chromatin biology from KMT2A::MLLT3 systems, and this beat was dispatched partly to ask whether that import is safe for H3K79me2 dynamics.

The honest answer from this deposit is: for the ORDER, yes; for the LEVEL, unknown. The per-gene ordering of H3K79me2 loss transfers between the two lines at a Spearman of 0.75-0.79, which is near the ceiling their shared baselines allow. So a statement of the form 'gene A loses more than gene B under DOT1L inhibition' imported from MOLM13 is likely to hold in RS4;11. A statement of the form 'X% is retained' is not, and this deposit shows it swinging by 2.5x in either direction depending on the compound.

That is a narrower licence than the corpus has been using, and it is the one the data supports. 1

Facts and figures
scriptanalyses/absolute-h3k79me2-in-the-index-fusion/run.py
statusrun
result summaryEPZ d9: RS4;11 3.93% vs MOLM13 1.60% retained, 2.53x, positive in all 10 deciles. compound 10 d9: 1.70% vs 2.46%, 0.71x, sign reverses in deciles 5-10. Per-gene Spearman 0.785 and 0.751 against a baseline Spearman ceiling of 0.815.
n shared tss entries19214
n used16636
n dropped baseline filter2578
PCT dropped13.42
spearman between lines epz0.785
spearman between lines comp100.751
spearman of baselines0.815
prereg branchnone of C1/C2/C3 - logged as a pre-registration failure
inputs
[
  "GSE149183 RS4_11 and MOLM13 sitepro tables"
]
How this could bite us back
  • If a chromatin pharmacodynamic endpoint for a DOT1L inhibitor swings 2.5x in either direction between two KMT2A-rearranged lines depending on which inhibitor is used, then a PD threshold set in one line and one compound does not transfer to a trial population with mixed partners.
  • The corpus-wide risk this quantifies: importing an MLLT3 number as a KMT2A::AFF1 number is safe for ordering and unsafe for magnitude, and most of the load-bearing numbers in this graph are magnitudes.
Still open
  1. Is there any deposit anywhere with two KMT2A fusion partners, a spike-in and replicates, so a partner effect on H3K79me2 could actually be tested? This project has never had one.
  2. The per-gene rank agreement of 0.785 sits just under the baseline agreement of 0.815. Is the residual 0.03 real partner-specific structure, or is it the ceiling itself? With single libraries that cannot be answered, and it is the only place a partner effect could hide.
  3. Does the NSD2 E1099K allele in RS4;11 alter H3K79me2 turnover? H3K36me2 and H3K79me2 are both gene-body marks and an antagonism has been proposed; nobody has tested it in this line.
Every source behind this entry (2)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    In GSE149183, RS4;11 retains 2.53x more H3K79me2 than MOLM13 at day 9 under EPZ5676 in every baseline decile, and 0.71x under compound 10 with the sign reversing in deciles 5-10, while per-gene rank agreement between the lines is 0.75-0.79.

    GSE149183 series supplementary files: RS4_11 and MOLM13 H3K79me2 TSS-ChIPseq signal + dm6 spike-in sitepro tables · Perner F, Gadrey JY, Xiong Y, et al. · NCBI Gene Expression Omnibus 2020
    What this evidence cannot show

    One library per condition per line: a between-line difference is a difference of two single measurements and cannot be tested. The RS4;11 arm is restricted to the 19,214 TSS entries that are genuinely RS4;11, and that subset has a median baseline of 107.6 against 53.6 for the full annotation, so the shared-gene comparison is made on signal-enriched TSSs. 2,578 of 19,214 shared entries (13.4%) were dropped for a baseline below 25 in either line. Each line's absolute scale rests on its own single dm6 count, so the LEVEL comparison inherits two unreplicated scale factors while the RANK comparison inherits none.

    Evidence that argues against this (1)
    • The rank agreement of 0.785 is itself the strongest argument that the level difference is technical: if the partner genuinely changed which genes depend on DOT1L, the ordering should diverge, and it does not - it sits at the ceiling their baselines allow. Orlando DA et al. 2014
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    RS4;11 (CVCL_0093) is a female 32Y B-ALL line with in-frame KMT2A::AFF1 and an NSD2 p.Glu1099Lys mutation; MOLM-13 (CVCL_2119) is a male 20Y acute monocytic leukaemia line with KMT2A::MLLT3.

    What this evidence cannot show

    Fetched from the Cellosaurus REST API 2026-09-23. NSD2 E1099K is recorded there as heterozygous, citing PMID 35124168 and DepMap ACH-000874.

    Evidence that argues against this (1)
    • NSD2 E1099K is a gain-of-function H3K36 dimethyltransferase allele, so RS4;11 carries a second chromatin lesion that MOLM13 does not. Any between-line chromatin difference has this as an alternative explanation before the fusion partner is reached. Cellosaurus, SIB 2026
Connected to