start / graph / ana-q130-the-eviction-gradient-exists-in-h3k79me2-and-is-the-steepest-of-any-active-mark-in-these-bins

AnalysismoderateQ130-does-the-eviction-gradient-exist-in-h3k79me2

In murine MLL-AF9, H3K79me2's response to a menin inhibitor runs 0.586 / 0.819 / 0.928 / 1.187 across the four eviction-depth quartiles of the fusion's own top-0.1% sites - Spearman -1.000, span 0.6005, the largest of any mark measured in these bins except H3K27me3 - while the aggregate over the same 1,228 bins is a single unremarkable 0.8156

also called the H3K79me2 eviction gradient · Q130's headline · the active mark leaves in proportion to how much fusion left

When a menin inhibitor pulls the leukaemia fusion protein off a stretch of DNA, the activating chemical tag DOT1L leaves behind disappears from that stretch in proportion to how much of the fusion left. Averaging over all the stretches at once hides this completely, because the places that lose the most and the places that gain are counted together.

The same thing, for a scientist

**Deposited data, re-analysed; no wet work and zero downloads** - every track was fetched and binned by Q124 in cycle 83 and was read off disk [E1]. **In murine MLL-AF9 (KMT2A::MLLT3) myeloid cells**, GSE260742's NTvKO control culture, 500 nM VTP-50469 for 48 h against DMSO 72 h, scored at `S_FLAG` (the FLAG-DMSO top 0.1% of `U_m`, 1,228 bins) split into Q117's four eviction-depth quartiles: | | Q1 most fusion lost | Q2 | Q3 | Q4 least | Spearman vs depth | span | aggregate | |---|---:|---:|---:|---:|---:|---:|---:| | **H3K79me2 VTP/DMSO, control genotype** | **0.5860** | 0.8188 | 0.9280 | **1.1865** | **-1.000** | **0.6005** | 0.8156 | | H3K79me2 VTP/DMSO, PCGF1-KO clone | 0.6123 | 0.7955 | 0.8928 | 1.0464 | **-1.000** | 0.4341 | 0.7990 | **At the bins the drug strips the most fusion from, H3K79me2's share falls by 41.4%. At the bins it strips the least, H3K79me2's share RISES by 18.7%.** The aggregate 0.8156 - the number Q124 committed in cycle 83 and the number the whole 'is a resistant cell drug-engaged?' argument was run on - is those two facts averaged into one [E1][E2]. **This is the largest span of any mark yet measured in these bins except H3K27me3.** In the same 1,228 bins, same culture, same drug, same normalisation: H3K27me3 1.0227, **H3K79me2 0.6005**, H2AK119ub 0.3820 (paired) and 0.1567 (rep2), H3K4me3 0.2534 [E3]. **It is not regression to the mean.** Each quartile was matched to bins outside `S_FLAG` on the decile of the same track's own DMSO share, at the same quartile sizes, 20 draws, seed 20260923 - Q125's construction and Q125's seed. The matched null is flat at 1.0301 / 1.0398 / 1.0566 / 1.0528, **no null draw of 20 reached the observed span**, and the residual observed-minus-null keeps Spearman -1.000 [E4]. **And it is not a property of a handful of bins.** The MEDIAN per-bin log2 share ratio is itself monotone at Spearman -1.000 (-0.7213 / -0.1680 / -0.0158 / +0.2673), the fraction of bins individually below 1 falls from 89.3% at Q1 to 24.7% at Q4, the bin-level bootstrap 95% interval for the span is [-0.651, -0.545], and dropping the single most influential bin moves the span by 0.0057 [E5]. **And it reproduces across every murine VTP contrast this deposit holds.** A replicate pair for a second culture was found already binned on disk during this beat, at zero download cost: **all four murine VTP-50469 H3K79me2 contrasts give Spearman -1.000**, with spans -0.2438, -0.4847, -0.6005 and -0.4341. The sign and the monotonicity reproduce four times out of four; the magnitude reproduces to within a factor of 1.99 between two replicates of one culture, and that factor is the honest error bar on every span quoted in this beat [E7]. **And it holds in a human cell, in a different deposit, with a different stratifier.** In MOLM13 (human KMT2A::MLLT3 AML) from GSE127507 - the VTP-50469 founding paper's own series - stratified by MLL1-N eviction depth, H3K79me2 runs **0.7746 / 1.0601 / 1.0995 / 1.1421**, Spearman **-1.000**, span 0.3675, **against an aggregate over the same sites of 0.998**. Stratified by MENIN eviction instead: 0.6437 / 0.6650 / 0.7061 / 0.9005, Spearman -1.000. In that series the INPUT is flat (Spearman +0.400, span 0.012, p = 0.90), which is the negative control the murine arm could not supply [E8].

Go deeper

Why the aggregate is not wrong, and is still the wrong number

0.8156 is arithmetically correct. It is the ratio of two sums over a set that spans a 3.5-fold range of eviction depth, and the quantity being summed is monotone in that depth. Averaging a monotone function over the range it is monotone across returns its mean, and the mean here is a number that belongs to no bin.

This is the second time the same shape has appeared in this corpus in two cycles. Q125 found it for H2AK119ub and H3K27me3 and wrote that S_FLAG is a set defined by where the fusion is and not by where the drug takes it away. H3K79me2 is the third mark and the sharpest case, because its aggregate is not near 1.0 - it is 0.8156, which reads as a real, modest, uninteresting effect, and is in fact a 41% loss and a 19% gain in the same set.

What the sign means, and the prediction that was written down first

The pre-registration (section 6, G2) predicted -1.000, on the grounds that H3K79me2 is DOT1L's product, DOT1L travels with the MLL-AF9 fusion, and the other active mark in these bins - H3K4me3 - already runs at -1.000. It also stated that +1.000 would be read as an instrument alarm first and a finding second, because an active fusion-borne mark behaving like the repressive marks is biologically incoherent.

The observed sign is negative. That is the predicted direction, and the prediction is on disk in analyses/q130-does-the-eviction-gradient-exist-in-h3k79me2/PREREGISTRATION.md before any H3K79me2 quartile number existed.

What this does NOT say

Not abundance. Every figure is a within-sample share ratio. Nothing here says H3K79me2 went up or down in total amount; it says its distribution moved. PMID 27875550 is the worked demonstration that this class of normalisation is blind to exactly that.
Not human, not B-lineage, not the index fusion. Murine MLL-AF9 (KMT2A::MLLT3) myeloid blood cells with a retrovirally over-expressed FLAG-tagged construct.
Replicated in sign, not in magnitude. The NTvKO arm deposits one H3K79me2 ChIP per genotype per drug, so ITS bootstrap interval measures bin sampling inside one library. A replicate pair for a different culture was found already binned on disk and scored (DEVIATION 4): all four murine VTP-50469 contrasts in this deposit give Spearman -1.000, with spans -0.2438 / -0.4847 / -0.6005 / -0.4341, so the span reproduces only to within a factor of 1.99 7.
Not a full 72-hour effect. VTP is 48 h against a 72 h vehicle, so 0.5860 is a lower bound on the loss and 1.1865 a lower bound on the gain.

Facts and figures

script	analyses/q130-does-the-eviction-gradient-exist-in-h3k79me2/run.py
status	run
result summary	H3K79me2 VTP/DMSO at S_FLAG's four eviction quartiles is 0.5860 / 0.8188 / 0.9280 / 1.1865 in the control genotype (Spearman -1.000, span 0.6005) against an aggregate of 0.8156; matched null flat at 1.030-1.057 with 0 of 20 draws reaching the span.
empirical p span	0.0476
absence class	measured-not-analysed
absence class note	Every track used here was public from 2025-11-23, was downloaded and binned by Q124 on 2026-09-23, and has sat on disk in C:/Temp/q124 since. Q117's eviction quartiles have been committed since cycle 68. The stratification cost zero downloads and about four minutes of compute; it had never been run because Q124's question was about a genotype and Q125's arm did not include H3K79me2.
absence class assigned	2026-09-23 (Q130)
epigenetic readout	H3K79me2 (DOT1L product) within-sample share ratio under a menin inhibitor, stratified by fusion-eviction depth; also H3K4me3, H3K27me3 and H2AK119ub in the same bins for comparison

aggregates S FLAG

span panel same bins same drug

inputs

[
  "GSM8123438 NTvKO Control - DMSO H3K79me2 72 h",
  "GSM8123439 NTvKO Control - SGC0946 5 uM H3K79me2 72 h",
  "GSM8123440 NTvKO Control - VTP-50469 500 nM H3K79me2 48 h",
  "GSM8123435/436/437 the same three in the PCGF1-KO clone C1",
  "GSM8123416 / GSM8123415 matched inputs",
  "GSM8123359 FLAG-DMSO and GSM8123361 FLAG-VTP (Q117, define S_FLAG and its quartiles)",
  "GSM8549823 input batch 2 (Q117, defines U_m)"
]

universe

{
  "grid_bins": 1316897,
  "blacklist_intervals_used": 3416,
  "bins_in_universe": 1227708,
  "k_top_0.1pct": 1228,
  "S_FLAG_n": 1228,
  "n_per_quartile": [
    308,
    306,
    307,
    307
  ],
  "quartile_cuts_read_from_q117_json": [
    0.0992,
    0.1748,
    0.2717
  ],
  "n_outside_S_FLAG": 1226480,
  "q117_expected_universe": null,
  "q124_expected_universe": 1227708,
  "q124_expected_k": 1228,
  "universe_reproduces_q124": true,
  "quartile_n_note": "Q117's own table reports 307/307/307/307 by rank; applying the stored CUTS as thresholds gives the numbers above. Q125's deviation 2 got the identical split, so the two beats' quartiles are the same bins."
}

per quartile VTP

{
  "control_NT": [
    0.586,
    0.8188,
    0.928,
    1.1865
  ],
  "pcgf1_ko_C1": [
    0.6123,
    0.7955,
    0.8928,
    1.0464
  ]
}

per quartile SGC

{
  "control_NT": [
    0.1396,
    0.1453,
    0.1452,
    0.136
  ],
  "pcgf1_ko_C1": [
    0.1334,
    0.1325,
    0.1349,
    0.1225
  ]
}

aggregates S FLAG — the numbers behind the chart

{
  "NT_VTP": 0.8156,
  "C1_VTP": 0.799,
  "NT_SGC": 0.1415,
  "C1_SGC": 0.1314
}

matched null median NT VTP

[
  1.0301,
  1.0398,
  1.0566,
  1.0528
]

residual NT VTP

[
  -0.4441,
  -0.221,
  -0.1286,
  0.1337
]

median per bin log2 NT VTP

[
  -0.7213,
  -0.168,
  -0.0158,
  0.2673
]

bootstrap span ci95 NT VTP

[
  -0.651,
  -0.5446
]

span panel same bins same drug — the numbers behind the chart

{
  "K119_PAIRED_NT_VTP": 0.382,
  "K119_REP2_NT_VTP": 0.1567,
  "K27_NT_VTP": 1.0227,
  "K4_NT_VTP": -0.2534,
  "K79_NT_VTP": -0.6005
}

all four murine vtp contrasts

{
  "rep1_span": -0.2438,
  "rep2_span": -0.4847,
  "absolute_difference": 0.2409,
  "ratio_larger_over_smaller": 1.988,
  "all_four_VTP_spans_in_this_deposit": {
    "MLLAF9 Control rep1 (GSM8123374)": -0.2438,
    "MLLAF9 Control rep2 (GSM8123388)": -0.4847,
    "NTvKO Control (GSM8123440)": -0.6005,
    "NTvKO PCGF1-KO (GSM8123437)": -0.4341
  },
  "all_four_VTP_spearmans": {
    "MLLAF9 Control rep1": -1,
    "MLLAF9 Control rep2": -1,
    "NTvKO Control": -1,
    "NTvKO PCGF1-KO": -1
  },
  "all_four_SGC_spans": {
    "MLLAF9 Control rep1": 0.0031,
    "MLLAF9 Control rep2": 0.0088,
    "NTvKO Control": 0.0036,
    "NTvKO PCGF1-KO": 0.0109
  }
}

human independent series MOLM13

{
  "series": "GSE127507, SubSeries of GSE127508, PMID 31821784 (Krivtsov 2019, the VTP-50469 founding paper)",
  "line": "MOLM13, human KMT2A::MLLT3 (MLL-AF9) AML",
  "MLL1n_stratified_H3K79me2": {
    "R_by_quartile_Q1_to_Q4": [
      0.7746,
      1.0601,
      1.0995,
      1.1421
    ],
    "spearman_vs_eviction_depth": -1,
    "span_Q1_minus_Q4": -0.3675,
    "R_aggregate_over_S": 0.998,
    "median_per_bin_log2_by_quartile": [
      -0.1019,
      0.116,
      0.159,
      0.215
    ],
    "bootstrap_span_ci95": [
      -0.407,
      -0.3266
    ],
    "matched_null_median_by_quartile": [
      0.9392,
      0.9463,
      0.9442,
      0.9367
    ],
    "null_span_max_abs": 0.053,
    "n_null_draws_reaching_observed_span": 0,
    "empirical_p_span": 0.0476
  },
  "MENIN_stratified_H3K79me2": {
    "R_by_quartile_Q1_to_Q4": [
      0.6437,
      0.665,
      0.7061,
      0.9005
    ],
    "spearman_vs_eviction_depth": -1,
    "span_Q1_minus_Q4": -0.2568,
    "R_aggregate_over_S": 0.7107,
    "median_per_bin_log2_by_quartile": [
      -0.8932,
      -0.849,
      -0.4182,
      -0.0356
    ],
    "bootstrap_span_ci95": [
      -0.3052,
      -0.2091
    ],
    "matched_null_median_by_quartile": [
      0.9565,
      0.9466,
      0.9373,
      0.9429
    ],
    "null_span_max_abs": 0.0465,
    "n_null_draws_reaching_observed_span": 0,
    "empirical_p_span": 0.0476
  },
  "MLL1n_stratified_INPUT_negative_control": {
    "R_by_quartile_Q1_to_Q4": [
      0.9497,
      0.974,
      0.9603,
      0.9377
    ],
    "spearman_vs_eviction_depth": 0.4,
    "span_Q1_minus_Q4": 0.012,
    "R_aggregate_over_S": 0.9542,
    "median_per_bin_log2_by_quartile": [
      -0.0704,
      -0.0222,
      -0.0519,
      -0.062
    ],
    "bootstrap_span_ci95": [
      -0.0222,
      0.0449
    ],
    "matched_null_median_by_quartile": [
      0.9413,
      0.9311,
      0.9172,
      0.9004
    ],
    "null_span_max_abs": 0.0642,
    "n_null_draws_reaching_observed_span": 18,
    "empirical_p_span": 0.9048
  }
}

How this could bite us back

If the menin inhibitor's H3K79me2 effect is concentrated at the most-evicted sites, then any pharmacodynamic assay that averages over a fusion-occupancy set will under-read engagement in a good responder by the same arithmetic that produced 0.8156. A dose-finding or target-engagement decision made on the unstratified number is made on a statistic that is insensitive by construction.
The gradient means the readout is shaped by how much fusion the drug removed, so a partially penetrating dose sits in the Q3/Q4 regime where H3K79me2 barely falls and may even rise. A single-threshold assay reads a partial responder and a non-responder alike.
H3K79me2 loss is also what a DOT1L inhibitor does, far more strongly. Any clinical assay built on H3K79me2 has to distinguish the two, and it cannot do so on magnitude - only on the eviction-depth shape, which requires a paired fusion-occupancy measurement in the same sample. That is a hard requirement for a patient assay and may be impractical.

Still open

PARTLY ANSWERED IN THIS BEAT, AND THE REMAINING HALF IS THE LINEAGE. The human question was closed by DEVIATION 5: in MOLM13 (human KMT2A::MLLT3) from the independent series GSE127507, stratified by MLL1-N eviction depth, H3K79me2 runs 0.7746 / 1.0601 / 1.0995 / 1.1421 at Spearman -1.000 against an aggregate of 0.998. WHAT IS STILL OPEN is the index fusion and the lineage: every system with this measurement is KMT2A::MLLT3 myeloid, and this graph exists for infant B-lineage KMT2A::AFF1. No series with an H3K79me2 arm, a menin inhibitor and a matched fusion-occupancy measurement in a KMT2A::AFF1 cell has been identified.
Where does the displaced share go? H3K79me2's share falls 41% at Q1 and RISES 19% at Q4, and Q124 measured the mid-occupancy floor falling to 0.723 under the same drug. A within-sample share sums to one, so the mark must accumulate somewhere this beat did not look.
Does the gradient track transcription? No RNA-seq from these exact samples has been located, so 'the mark leaves' is a statement about chromatin and not about output.
Is the Q4 gain real or is it the arithmetic shadow of the Q1 loss? A spike-in-normalised H3K79me2 arm would separate them and none exists in this series.

Every source behind this entry (8)

E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
In murine MLL-AF9 (GSE260742 NTvKO control, cell line MLLAF9, myeloid, blood, taxid 10090), 500 nM VTP-50469 48 h against DMSO 72 h gives H3K79me2 within-sample share ratios of 0.5860 / 0.8188 / 0.9280 / 1.1865 across Q117's four eviction-depth quartiles of S_FLAG (Spearman -1.000 against eviction depth, span 0.6005) against an aggregate of 0.8156 over the same 1,228 bins.
Q130 re-analysis of GSE260742: the per-quartile H3K79me2 profile at S_FLAG, both genotypes, both drugs. analyses/q130-does-the-eviction-gradient-exist-in-h3k79me2/run.py, deviation1_per_bin_robustness.py, deviation2_two_criteria_i_wrote_badly.py, PREREGISTRATION.md and results.json, run 2026-09-23. Zero downloads: every track was binned by Q124 and read from disk. · Iqbal Lab oncofusion knowledge graph 2026
What this evidence cannot show
NO SPIKE-IN NORMALISATION IS POSSIBLE. Q117 verified zero hits for the entire spike-in vocabulary across all 419,824 characters of GSE260742's 158 sample records and no scale factor is deposited, so every cross-condition number here is a WITHIN-SAMPLE SHARE. It cancels the unknown per-sample scale factor exactly and IS BLIND TO A UNIFORM CHANGE BY CONSTRUCTION: nothing here measures H3K79me2's abundance, everything measures its distribution. EXPOSURE ASYMMETRY: DMSO 72 h and SGC0946 5 uM 72 h are matched, VTP-50469 500 nM is 48 h, so every VTP ratio is a LOWER BOUND on the 72-hour effect - identical in both genotypes, so the genotype DIFFERENCE is unaffected. REPLICATION, CORRECTED MID-BEAT: the NTvKO arm deposits ONE H3K79me2 ChIP per genotype per drug, so ITS spans carry no technical error bar and its bin-level bootstrap intervals measure sampling within one library, NOT library-to-library variation. A REPLICATE PAIR DOES EXIST for a different culture - GSM8123372/374 and GSM8123386/388, both titled MLLAF9 Control, the second carrying `rep2` - already binned by Q122 and scored in Q130 DEVIATION 4: the SIGN and the monotonicity reproduce (Spearman -1.000 in all four murine VTP contrasts in this deposit) while the SPAN reproduces only to within a factor of 1.99 (-0.2438 against -0.4847). EVERY SPAN IN THIS BEAT SHOULD BE READ WITH THAT 2-FOLD REPRODUCIBILITY. THE KNOCKOUT is one CRISPR guide, one clone C1, no second clone, no rescue, and NO deposited PCGF1 protein or transcript measurement. S_FLAG is defined by a DYKDDDDK-tag ChIP on a RETROVIRALLY OVER-EXPRESSED tagged MLLAF9 construct, so it is the over-expressed molecule's sites and not an endogenous fusion's. THE H3K79me2 ANTIBODY IS NOT RESOLVED - `antibod` occurs zero times in the 158 GEO records - and is recorded unresolved rather than guessed. SYSTEM: murine MLL-AF9 (KMT2A::MLLT3) myeloid blood cells, NOT the infant B-lineage KMT2A::AFF1 disease this graph is about.
Evidence that argues against this (2)
- A published methods study establishes that STANDARD ChIP-seq normalisation FAILS TO DETECT a genome-wide decrease in a histone mark under an inhibitor of its writer: standard methods see no H3K27me3 reduction under an EZH2 inhibitor, and a substantial reduction appears only with a Drosophila chromatin spike-in. Every number in this entry is exactly the kind of normalisation that paper shows to be blind. Egan B et al. 2016
- The system is murine MLL-AF9 (KMT2A::MLLT3) myeloid cells carrying a retrovirally over-expressed FLAG-tagged construct. This graph exists for human infant B-lineage KMT2A::AFF1, and this corpus's commonest documented synthesis failure is generalising a murine MLL-AF9 myeloid result to that disease. Neville D et al. 2026
E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
Q124 committed the aggregate figure in cycle 83: V_NT = 0.8156 and V_KO = 0.7990 for H3K79me2 at S_FLAG, and scored H3K79me2 only as an aggregate - no per-quartile H3K79me2 ratio appears anywhere in its results.json.
Q124 re-analysis of GSE260742, committed cycle 83: arms.arm2_pcgf1_knockout.ratios.S_FLAG and headline.arm2. analyses/q124-the-timecourse-and-the-pcgf1-knockout/results.json, read from the committed JSON rather than retyped. · Iqbal Lab oncofusion knowledge graph 2026
What this evidence cannot show
Read out of the committed JSON in code, not retyped. Q124's scope decision was correct for its own question, which was whether the knockout changes the drug response; the stratification is a different question and Q117's quartiles were already on disk when Q124 ran.
Evidence that argues against this (1)
- Q124's aggregate is not refuted by this entry and the arithmetic is unchanged. What is refuted is the reading of 0.8156 as a description of what happens at the fusion's sites. Iqbal Lab oncofusion knowledge graph 2026
E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Across the same 1,228 S_FLAG bins, the same culture and the same drug, the Q1-to-Q4 span is 1.0227 for H3K27me3, 0.6005 for H3K79me2, 0.3820 and 0.1567 for the two H2AK119ub conventions and 0.2534 for H3K4me3, so H3K79me2 carries the second-largest span of the five contrasts and the largest of any active mark.
Q125 re-analysis of GSE260742, committed cycle 88: the per-quartile H2AK119ub, H3K27me3 and H3K4me3 profiles, read from the committed deviation2 JSON. analyses/q125-the-h2ak119ub-arm-of-the-pcgf1-knockout/deviation2-quartiles-and-conventions.json and deviation3-matched-null.json, read from the committed JSON rather than retyped. · Iqbal Lab oncofusion knowledge graph 2026
What this evidence cannot show
The H2AK119ub PAIRED arm is .filt.bam.tdf while H3K79me2, H3K27me3, H3K4me3 and the H2AK119ub REP2 arm are bare .tdf; both H2AK119ub conventions are therefore quoted rather than one being chosen. A span is not an effect size in any absolute sense - it is the difference between two ratios of within-sample shares - and comparing spans across marks assumes the marks' dynamic ranges are comparable, which is not established.
Evidence that argues against this (1)
- The two H2AK119ub conventions disagree by more than a factor of two in span (0.3820 against 0.1567), which is Q125's own finding that two replicates of one culture disagree in sign and span. Iqbal Lab oncofusion knowledge graph 2026
E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
A DMSO-decile-matched null drawn from outside S_FLAG at the same quartile sizes gives a flat 1.0301 / 1.0398 / 1.0566 / 1.0528, none of 20 draws reached the observed span of 0.6005 (empirical two-sided p = 1/21 = 0.0476), and the residual observed-minus-null retains Spearman -1.000.
Q130 re-analysis of GSE260742: the DMSO-decile-matched null, 20 draws carried through as complete four-quartile vectors, seed 20260923. analyses/q130-does-the-eviction-gradient-exist-in-h3k79me2/run.py, deviation1_per_bin_robustness.py, deviation2_two_criteria_i_wrote_badly.py, PREREGISTRATION.md and results.json, run 2026-09-23. Zero downloads: every track was binned by Q124 and read from disk. · Iqbal Lab oncofusion knowledge graph 2026
What this evidence cannot show
With 20 draws the smallest attainable p is 0.0476, so the p is limited by the null's size and not by the effect; it should be read as 'no draw came close' rather than as a precise significance. THE NULL DRAWS ARE NOT INDEPENDENT of the observed data in one respect: they are drawn from the same libraries, which is the point (they share every library artefact) but means they cannot bound a defect that is specific to S_FLAG bins. 3 of the 20 null draws happened to produce |Spearman| = 1.000, against a chance expectation of 1.67, which is why the span and not the Spearman is the scored statistic.
Evidence that argues against this (1)
- A matched null built from the same libraries cannot detect an artefact that is specific to the S_FLAG bins themselves, for instance a systematic mappability or copy-number feature of the fusion's strongest sites. Iqbal Lab oncofusion knowledge graph 2026
E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
The gradient is a per-bin fact: the median per-bin log2 share ratio across the four quartiles is -0.7213 / -0.1680 / -0.0158 / +0.2673 (Spearman -1.000), the fraction of bins individually below 1 falls 89.3% to 24.7%, the bin-level bootstrap 95% interval for the span is [-0.651, -0.545] and excludes zero in 2,000 of 2,000 resamples, and leave-one-bin-out changes the span by at most 0.0057.
Q130 re-analysis of GSE260742: DEVIATION 1, per-bin robustness: median per-bin log2 ratio, fraction below 1, 2,000-resample bin-level bootstrap and leave-one-bin-out. analyses/q130-does-the-eviction-gradient-exist-in-h3k79me2/run.py, deviation1_per_bin_robustness.py, deviation2_two_criteria_i_wrote_badly.py, PREREGISTRATION.md and results.json, run 2026-09-23. Zero downloads: every track was binned by Q124 and read from disk. · Iqbal Lab oncofusion knowledge graph 2026
What this evidence cannot show
A bin-level bootstrap resamples bins within one library. It measures how much of the span is attributable to which bins were chosen and says NOTHING about library-to-library variation, which is unmeasurable here because no replicate exists. POST HOC: none of these four statistics is in the pre-registration.
Evidence that argues against this (1)
- A median per-bin ratio and a ratio of sums are different statistics and can disagree; agreement here is reassuring but is not independent evidence, because both are computed from the same two tracks over the same bins. Iqbal Lab oncofusion knowledge graph 2026
E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
VTP-50469 is a menin-KMT2A interaction inhibitor whose effect on MLL-rearranged leukaemia chromatin was established in the founding paper this project's whole eviction model rests on.
A Menin-MLL Inhibitor Induces Specific Chromatin Changes and Eradicates Disease in Models of MLL-Rearranged Leukemia. · Krivtsov AV, Evans K, Gadrey JY, et al. · Cancer Cell 2019
PMID 31821784 DOI
What this evidence cannot show
Cited for the identity and mechanism of the compound, not for anything measured in this entry. That paper's own system is human and murine MLL-rearranged leukaemia, not the GSE260742 cultures re-analysed here.
Evidence that argues against this (1)
- A drug's named mechanism is not evidence that a chromatin change observed under it is caused by that mechanism. Iqbal Lab oncofusion knowledge graph 2026
E7moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
All four murine VTP-50469 H3K79me2 contrasts deposited in GSE260742 - MLLAF9 Control (GSM8123374), MLLAF9 Control rep2 (GSM8123388), NTvKO Control (GSM8123440) and NTvKO PCGF1-KO (GSM8123437) - give Spearman -1.000 against eviction depth, with spans of -0.2438, -0.4847, -0.6005 and -0.4341, so the sign reproduces four times out of four and the magnitude reproduces to within a factor of 1.99 between two replicates of one culture.
Q130 re-analysis of GSE260742: DEVIATION 4, the replicate pair already binned by Q122 in cycle 82 and scored here at zero download cost. analyses/q130-does-the-eviction-gradient-exist-in-h3k79me2/run.py, deviation1_per_bin_robustness.py, deviation2_two_criteria_i_wrote_badly.py, PREREGISTRATION.md and results.json, run 2026-09-23. Zero downloads: every track was binned by Q124 and read from disk. · Iqbal Lab oncofusion knowledge graph 2026
What this evidence cannot show
THE THIRD-CULTURE ARGUMENT DOES NOT APPLY to the two MLLAF9 Control contrasts, because S_FLAG is defined in that same culture (GSM8123359). They are the better REPLICATE and the worse CONTROL; the NTvKO contrasts are the reverse; the four together are quoted precisely because no single one has both properties. The rep1 arm is .filt.bam.tdf and the rep2 arm is bare .tdf, so the pair is not a pure technical replicate. AND THE PAIR'S STATUS IS DISPUTED: the rep2 samples' GEO titles say 'rep2' while their supplementary filenames say 'F3_NT'. The companion entry reports that discrepancy rather than resolving it, and either reading - a replicate or a second culture - supports this claim, which is why it is made before the dispute is settled.
Evidence that argues against this (1)
- A span that reproduces only to within a factor of two is a weak quantitative measurement, and this entry's headline quotes the largest of the four spans. Iqbal Lab oncofusion knowledge graph 2026
E8moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
In MOLM13, a HUMAN KMT2A::MLLT3 AML line, from the INDEPENDENT series GSE127507 (SubSeries of GSE127508, PMID 31821784), 330 nM VTP-50469 for 3 days against DMSO gives H3K79me2 share ratios of 0.7746 / 1.0601 / 1.0995 / 1.1421 across MLL1-N eviction-depth quartiles - Spearman -1.000, span 0.3675, 0 of 20 matched-null draws reaching it - while the aggregate over the same 1,389 sites is 0.998.
Q130 re-analysis of GSE260742: DEVIATION 5, the human independent-series test on GSE127507 MOLM13 tracks binned by Q120. analyses/q130-does-the-eviction-gradient-exist-in-h3k79me2/run.py, deviation1_per_bin_robustness.py, deviation2_two_criteria_i_wrote_badly.py, PREREGISTRATION.md and results.json, run 2026-09-23. Zero downloads: every track was binned by Q124 and read from disk. · Iqbal Lab oncofusion knowledge graph 2026
What this evidence cannot show
FIVE WEAKNESSES RELATIVE TO THE MURINE TEST, all stated in the deviation's header. (1) No tagged fusion: MLL1-N is a SHARED epitope that sees wild-type KMT2A as well as the fusion, and MENIN is the drug's own target. (2) NO THIRD CULTURE - the stratifier and the measurement are the same MOLM13 libraries, so a shared-library artefact is not structurally excluded here. (3) No SGC0946 arm in this series, so the drug-shape control does not exist in it. (4) Quartiles by RANK, because Q117's murine cut values have no meaning on a human MLL1n ratio. (5) n = 1 per antibody per condition. THE SERIES ALSO RAN A DROSOPHILA SPIKE-IN, MEASURED IT AND DELIBERATELY DID NOT USE IT, and no scale factor is deposited, so this arm is a within-sample share like every other.
Evidence that argues against this (1)
- MLL1-N cannot distinguish the fusion from wild-type KMT2A, and cycle 83 recorded that the MLL1-N channel moves as much under a DOT1L inhibitor as under a menin inhibitor (0.3012 against 0.3298), so a stratifier built on it is not a clean fusion-eviction measure. Iqbal Lab oncofusion knowledge graph 2026

Connected to

related to →DOT1L — the target where excellent chemistry met an indirect interaction
H3K79me2 is DOT1L's product; the gradient is in DOT1L's mark
related to →KMT2A::MLLT3 (MLL-AF9), t(9;11)(p21.3;q23.3)
murine MLL-AF9, the fusion whose eviction defines the strata
depends on →MEN1 / menin — the adaptor the fusion cannot work without
the eviction is produced by inhibiting the menin-KMT2A interaction
supports →At the 2 kb bins a menin inhibitor strips the most fusion from, both repressive marks GAIN and the active mark is LOST, monotonically in eviction depth (Spearman +1.000, +1.000 and -1.000) against a baseline-matched null that is flat to within 0.007 - the silencing clause of hyp-q117's prediction 5, measured for the first time
the fourth mark in the same 1,228 bins, and the one Q125 named as the highest-value item left
← part ofThe menin inhibitor's effect on H3K79me2 is a REDISTRIBUTION within the fusion's own site set, not a loss from it: the share falls 41.4% at the quartile the fusion is most evicted from and RISES 18.7% at the quartile it is least evicted from, in murine MLL-AF9
the Q4 gain is one end of the gradient
← supportsWith H3K79me2 added, the eviction-depth panel over one set of 1,228 bins under one drug is H3K27me3 +1.000, H2AK119ub +1.000, H3K4me3 -1.000 and H3K79me2 -1.000 - a clean repressive-up / active-down split that no shared library, blacklist or regression artefact can produce
the sign-reversal control for the H3K79me2 gradient
← supportsSGC0946 removes 86% of H3K79me2's share at the fusion's own sites and the four eviction quartiles are 0.1396 / 0.1453 / 0.1452 / 0.1360 - span 0.0036, Spearman +0.400, a bootstrap interval spanning zero - so a very large effect on this mark can carry no eviction-depth structure whatever
the same-mark, same-bins drug control that makes the gradient interpretable
← supportsThe H3K79me2 eviction gradient against a DMSO-decile-matched null drawn from outside the fusion's sites at the same quartile sizes: the null is flat at 1.030-1.057, none of 20 draws reached the observed span of 0.6005, and the residual retains Spearman -1.000 - but the restricted-population trap is real and the baseline IS graded 2.2-fold
the baseline-matched null for the gradient
← related toStratifying by eviction depth does NOT rescue H3K79me2 as a resistance-detecting readout: at the most-evicted quartile the PCGF1-knockout and the control are 0.6123 against 0.5860, a difference of 0.0263, while the whole genotype effect sits at the LEAST-evicted quartile (-0.1401) - the opposite of eviction-graded
the same gradient, scored across genotypes: it exists in both and does not separate them at Q1
← supportsThe H3K79me2 eviction gradient is not a ratio of sums dominated by a few bins: the MEDIAN per-bin log2 share ratio is itself monotone at Spearman -1.000, the fraction of bins individually below 1 falls 89.3% to 24.7%, the bootstrap span interval is [-0.651, -0.545], and no single bin moves the span by more than 0.0057
the per-bin robustness of the headline gradient
← supportsIn the VTP-50469 timecourse Q124 declared uninterpretable, the eviction-gradient SPAN grows monotonically 0.104 / 0.198 / 0.214 / 0.335 at 4/8/16/24 h with Spearman -1.000 at every timepoint - while the aggregate it was suppressed for is non-monotone and perfectly anti-correlated with library complexity. The span is better behaved, and it is NOT clean.
the same gradient across four exposure times, supporting only
← related toA Spearman over four quartiles has 24 orderings and 11 attainable values: rho = +1.000 is 1/24 = 0.042, |rho| = 1.000 is 2/24 = 0.083, and a +-20% band around 0.80 contains exactly one attainable value - so a perfect monotone over four strata is a weak statistic and must never be scored alone
the statistical bound on this beat's headline Spearman
← supportsThe eviction quartiles are defined in the 'MLLAF9 Control' FLAG culture and the H3K79me2 they stratify is measured in the 'NTvKO Control' and 'NTvKO PCGF1 KO' cultures - zero shared libraries - so the commonest way a stratified analysis fools itself is structurally unavailable here
the structural control the matched null cannot supply
← related toThe four eviction quartiles are not equivalent stretches of genome: all three GSE260742 input tracks give a monotone increasing share across Q1 to Q4 at Spearman -1.000, 1.15 / 1.17 / 1.13-fold - small, real, in the same direction as the H3K79me2 gradient, and it cancels exactly in a within-genotype drug/vehicle ratio
the structural confound the gradient survives, and the reason G5 is uninformative
← supportsAll four murine VTP-50469 H3K79me2 contrasts in GSE260742 give Spearman -1.000 against eviction depth, and all four SGC0946 contrasts are flat: the sign reproduces four times out of four, the span only to within a factor of 1.99, and the replicate pair that establishes this was already binned on disk from cycle 82
four of four murine VTP contrasts, and the honest span error bar
← supportsIn MOLM13, human KMT2A::MLLT3, from the VTP-50469 founding paper's own series GSE127507, H3K79me2's aggregate response at MLL1-N's top sites is 0.998 - and the same sites split by eviction depth run 0.7746 / 1.0601 / 1.0995 / 1.1421 at Spearman -1.000, with the INPUT stratified identically staying flat
the human, independent-deposit replication of the gradient
← related toGSM8123386/387/388 carry the GEO title 'MLLAF9 Control - ... H3K79me2 rep2' and the supplementary filename 'ChIP_K79_F3_NT_*'. Two committed analyses treat them as a separate culture and one asserts 'FOUR INDEPENDENT MURINE CULTURES'. If the titles are right there are three, one of them replicated. REPORTED AS A DISPUTE, NOT APPLIED.
bounds how independent this beat's four murine contrasts are
← related toRS4;11 - the human KMT2A::AFF1 index-fusion line - has a deposited H3K79me2 pair, a MENIN pair, an MLL1-N pair, a DOT1L pair and an input under the same menin inhibitor in GSE127507. Every track needed to run this beat's test on the index fusion exists. The download is 1.99 GB measured by HEAD request, and nobody has run it.
the index-fusion test this beat could not run
← supportsA menin inhibitor's chromatin signature is organised by how much fusion it removed from each site, so every mark's response is a GRADIENT in eviction depth - and any readout that averages over a fusion-occupancy set is underpowered by construction, sometimes to exactly 1.000
the H3K79me2 arm of the generalised hypothesis
← related toOn the absolute scale the H3K79me2 gradient is real and it runs the other way from the share: the most heavily marked decile retains 0.8% and the least marked retains 31%, a 41-fold gradient - and every decile, in both lines and under both drugs, still loses
an absolute-scale analogue of the eviction gradient, in the index fusion's own line