Oncofusion

start / graph / ana-q116-knocking-down-both-kdm4-erasers-raises-h3k9me3-at-active-promoters-and-not-over-the-kmt2a-gene-body

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q116-h3k9-axis-at-the-locus

RUN, with its positive control: knocking down KDM4B and KDM4C together raises H3K9me3 at the +1 nucleosome of 2,000 active promoters by 35.7 percentile points, moves the KMT2A promoter's +1 nucleosome by the same 35.3, and moves the 90 kb KMT2A gene body 5.05 points the WRONG way - so the methyl-poor domain is not an erasure product

also called the eraser-knockdown arm · shKDM4B+shKDM4C at the KMT2A locus · the only KDM4 perturbation in any archive that reads this mark

Switching off both enzymes that erase the silencing tag makes the tag pile up at switched-on genes across the genome, exactly as published. It does not make it pile up over the KMT2A gene, so the tag is missing there for some other reason than being erased.

The same thing, for a scientist

**The question, and why this arm is worth more than a locus number.** `shCTL` (GSM4133820, GSM4133821) against `shKDM4B_shKDM4C` (GSM4133822, GSM4133823), all H3K9me3 ChIP-seq in MCF-7, all -E2. This is the **only eraser perturbation in any public archive whose readout is H3K9me3 occupancy**, and Q114 named it as an internal positive control it did not have. If removing both erasers raises H3K9me3 genome-wide but not at *KMT2A*, the methyl-poor domain there is not simply an erasure product [E1]. **THE POSITIVE CONTROL FIRES, and it reproduces a published figure.** The paper the deposit belongs to reports *"major H3K9me3 accumulation on +1 nucleosome after knockdown of KDM4B/4C"* (Extended Data Fig. 7p). Measured independently here, against a gene set and window geometry declared before the tracks were reopened: the median **+1 nucleosome** (TSS+1 to TSS+200) of the **2,000 highest-expressed MCF-7 protein-coding genes** moves from the **21.73rd** percentile of a 4,000-window active-plus-silent set under shCTL to the **57.43rd** under the double knockdown - **Δ = +35.71 percentile points**. At the **1 kb promoter window** the same comparison gives **Δ = −4.79**, i.e. nothing. **The effect is confined to the +1 nucleosome, which is exactly where the paper says it is** [E2]. **So the arm is a working instrument, and the locus answer means something.** **At the *KMT2A* promoter's +1 nucleosome, the locus moves like everything else.** shCTL 0.00 → shKD **35.25** percentile of the active-promoter set, **Δ = +35.25** against a genome-wide Δ of +35.71. *KMT2A*'s promoter is an ordinary KDM4 substrate [E2]. **Over the 90 kb *KMT2A* gene body, it moves the other way, and it is the most reproducible number in this beat.** Percentile of width-matched MCF-7 CTCF anchors: shCTL **7.60 and 7.70** (within-arm spread **0.10** points), shKD **2.56 and 2.64** (spread **0.08**). **Δ = −5.05** - downward. Against the 385 matched transcribed gene bodies: 18.18 and 20.26 → 3.12 and 7.27, **Δ = −14.03**. **Removing both erasers does not restore H3K9me3 over the *KMT2A* gene body; it leaves it relatively poorer** [E1]. **By the pre-registered rule the gene body DOES NOT MOVE**, because |Δ| = 5.05 is below the declared 10-point threshold and below the genome-wide median |Δ| of 24.42 - even though it exceeds both within-arm replicate spreads by fiftyfold. **The rule and the data disagree about this one interval, and the rule is what I wrote down first, so the rule is what is reported** [E1]. **And the exon-11 element does not move because it cannot: it reads an explicit 0.0000 in all four tracks**, before and after the knockdown.

Go deeper
Every interval, both arms, with the replicate spreads beside the effect

Percentile of width-matched MCF-7 CTCF summit windows. The spreads are the point - at 90 kb the two replicates of each arm agree to 0.1 percentile points; at 1-2 kb they disagree by up to 77.

intervalwidthshCTLshKDΔshCTL spreadshKD spreadpre-registered verdict
exon111460.00 / 0.000.00 / 0.000.000.000.00DOES NOT MOVE
ctcf_element1,04731.45 / 0.0020.15 / 17.70+3.2031.452.45DOES NOT MOVE
ctcf_summit_1kb1,00032.51 / 0.0021.15 / 18.57+3.6132.512.58DOES NOT MOVE
dcal_cut_1kb99932.54 / 87.720.00 / 0.00−60.1355.180.00MOVES (see below - do not believe it)
bcr_core7,46563.61 / 61.6956.56 / 0.09−34.321.9156.47DOES NOT MOVE
kmt2a_tss_2kb2,00077.17 / 0.0071.52 / 59.19+26.7777.1712.33DOES NOT MOVE
kmt2a_gene90,3767.60 / 7.702.56 / 2.64−5.050.100.08DOES NOT MOVE

dcal_cut_1kb returns MOVES and the verdict is not credible. The two shCTL replicates read 32.54 and 87.72 there - a 55.18-point disagreement between two controls - while both shKD replicates read 0.00. The rule requires |Δ| to exceed both within-arm spreads, and −60.13 does, by 5 points. The rule is missing a ceiling on the within-control spread. This is recorded as deviation D2 in the pre-registration and the threshold is not re-tuned; no reader should act on a verdict at an interval whose two controls differ by more than half the percentile range.

Genome-wide, at 146 bp, the arm is mostly noise - and that is a finding about the window, not about the knockdown

Over all 56,171 CTCF anchor windows at 146 bp:

statisticvalue
median |Δpercentile| between the arm means24.42
95th percentile of |Δ|60.36
99th percentile of |Δ|78.25
anchors with |Δ| ≥ 1056.4%
anchors whose |Δ| exceeds their own within-arm replicate spread9.85%
anchors rising in the knockdown49.1%

Only one anchor in ten moves by more than its own replicate noise. A median |Δ| of 24 percentile points between conditions sounds enormous and is almost entirely the sampling variance of a 146 bp window in a ChIP-seq track whose modal interval is tens of base pairs.

The Spearman structure says the same thing, and says it uncomfortably. Replicate-to-replicate over the 56,171 anchors: shCTL vs shCTL ρ = 0.3247, shKD vs shKD ρ = 0.2592, and the mean of the four across-condition pairs is 0.2898. The pre-registered ARM UNINFORMATIVE test asks whether within-condition ρ exceeds across-condition ρ; it passes on the shCTL pair (0.325 > 0.290) and FAILS on the shKD pair (0.259 < 0.290). So the arm clears its own bar by the narrowest margin available, and the two knockdown replicates resemble each other less than a knockdown resembles a control. That is why the genome-wide positive control in E2 - computed over a different window class - is the evidence that the arm works, and the 146 bp anchor statistics are not.

What this does and does not settle about the methyl-poor domain

Settled, in MCF-7: the H3K9me3 deficit over the KMT2A gene body is not maintained by KDM4B/KDM4C erasure. Remove both and the deficit deepens rather than fills. The domain is either actively kept clear by something else, or it is what a transcribed 90 kb gene body looks like - and ana-q116-the-methyl-poor-kmt2a-domain-against-385-length-and-expression-matched-gene-bodies says about half of it is the latter.

Not settled: whether KDM4A maintains it. The knockdown removes the weak eraser (KDM4B) and one strong one (KDM4C) and leaves KDM4A, which UniProt ranks as the strongest H3K9me3 demethylase of the four, entirely untouched. A null from a partial family knockdown does not exonerate the family.

Also not settled: whether the direction is real. −5.05 percentile points with 0.1-point replicate spreads looks like a solid small effect, but the two arms are two shRNA transductions with no spike-in, and HOMER depth normalisation means a genuine genome-wide rise in the mark would push every non-rising region down in percentile terms. A locus that fails to rise when the genome rises will read as falling. That is precisely what the +35.7-point genome-wide rise at active promoters predicts for a locus that does not participate, and it is the most likely explanation of the −5.05. It does not change the conclusion; it changes its wording from "the locus loses H3K9me3" to "the locus does not join in."

Facts and figures
scriptanalyses/q116-h3k9-axis-at-the-locus/run116.py, q2_eraser_arm.py and q2_promoter_control.py
prereganalyses/q116-h3k9-axis-at-the-locus/PREREGISTRATION.md, section 8 (Q2) and deviations D2 and D7
statusrun
run date2026-09-23
run byQ116-h3k9-axis-at-the-locus
complete runtrue
headlineThe methyl-poor KMT2A gene body is NOT an erasure product. Removing both KDM4 erasers raises H3K9me3 at active-promoter +1 nucleosomes by 35.7 percentile points and leaves the KMT2A gene body relatively poorer, not richer.
localizationchromatin at chr11q23.3 (the seven stored KMT2A intervals), against 56,171 MCF-7 CTCF anchor windows, 385 matched transcribed gene bodies, and 4,000 promoter windows of 2,000 active and 2,000 silent genes
subnuclear localizationnot measured
cell typeMCF-7 (ER+ breast adenocarcinoma). NOT haematopoietic, NOT KMT2A-rearranged.
cell statehormone-deprived, shRNA-transduced; shCTL against shKDM4B+shKDM4C. No damage arm.
epigenetic readoutH3K9me3 under double knockdown of its two KDM4 erasers
eraser of that markKDM4B (O94953) and KDM4C - both removed; KDM4A and KDM4D untouched
shown for fusionNO - the perturbation exists only in MCF-7, and no KDM4 perturbation with an H3K9me3 readout exists in any leukaemia cell in any archive
absence classmeasured-not-analysed
absence class noteThe four samples were deposited in 2019 and the arm had never been read at this locus. Q114 named it as the internal positive control it lacked and could not afford to run.
genome wide 146bp
frac_anchors_abs_delta_ex…0.099frac_anchors_abs_delta_exceeding_own_replicate_spread: 0.0985frac_anchors_rising_in_KD0.491frac_anchors_rising_in_KD: 0.491frac_anchors_abs_delta_ov…0.564frac_anchors_abs_delta_over_10: 0.5638median_abs_delta_percenti…24.4median_abs_delta_percentile: 24.417p95_abs_delta_percentile60.4p95_abs_delta_percentile: 60.358p99_abs_delta_percentile78.2p99_abs_delta_percentile: 78.249n_anchors56,171n_anchors: 56171
per interval delta percentile
dcal_cut_1kb-60.1dcal_cut_1kb: -60.13bcr_core-34.3bcr_core: -34.32kmt2a_gene-5.050kmt2a_gene: -5.05exon110.000exon11: 0ctcf_element3.200ctcf_element: 3.2ctcf_summit_1kb3.610ctcf_summit_1kb: 3.61kmt2a_tss_2kb26.8kmt2a_tss_2kb: 26.77
results
[
  "analyses/q116-h3k9-axis-at-the-locus/results.json",
  "analyses/q116-h3k9-axis-at-the-locus/results_q2_eraser_arm.json",
  "analyses/q116-h3k9-axis-at-the-locus/results_q2_promoter_control.json"
]
arms
{
  "shCTL": [
    "GSM4133820",
    "GSM4133821"
  ],
  "shKDM4B_shKDM4C": [
    "GSM4133822",
    "GSM4133823"
  ]
}
arm informativeness
{
  "spearman_shCTL_vs_shCTL": 0.3247,
  "spearman_shKD_vs_shKD": 0.2592,
  "mean_spearman_across_condition": 0.2898,
  "preregistered_test": "within-condition rho must exceed across-condition rho",
  "result": "PASSES on the shCTL pair (0.3247 > 0.2898), FAILS on the shKD pair (0.2592 < 0.2898)",
  "verdict": "ARM INFORMATIVE - but on the strength of the genome-wide +1 nucleosome positive control, not on this test"
}
positive control plus1 nucleosome
{
  "declared_before_reopening_the_tracks": true,
  "window": "TSS+1 to TSS+200 in the direction of transcription",
  "gene_sets": "2,000 highest-TPM protein-coding genes (ENCFF251DYL, TPM 63.74-10,242.87) versus 2,000 sampled from 5,644 with TPM < 0.1, seed 116",
  "shCTL_median_active_percentile": 21.725,
  "shKD_median_active_percentile": 57.431,
  "delta": 35.706,
  "per_replicate_shCTL": [
    43.45,
    0
  ],
  "per_replicate_shKD": [
    71.69,
    43.18
  ],
  "active_window_zero_fraction_per_track": [
    0.34,
    0.81,
    0.13,
    0.41
  ],
  "published_direction_reproduced": "PMID 34194047 Extended Data Fig. 7p",
  "same_comparison_at_the_1kb_promoter_window": -4.794,
  "reading": "the effect is confined to the +1 nucleosome, exactly where the paper places it"
}
kmt2a plus1 nucleosome
{
  "shCTL_percentile_of_active_set": 0,
  "shKD_percentile_of_active_set": 35.25,
  "delta": 35.25,
  "reading": "moves like an average active promoter"
}
genome wide 146bp — the numbers behind the chart
{
  "n_anchors": 56171,
  "median_abs_delta_percentile": 24.417,
  "p95_abs_delta_percentile": 60.358,
  "p99_abs_delta_percentile": 78.249,
  "frac_anchors_abs_delta_over_10": 0.5638,
  "frac_anchors_abs_delta_exceeding_own_replicate_spread": 0.0985,
  "frac_anchors_rising_in_KD": 0.491
}
per interval delta percentile — the numbers behind the chart
{
  "exon11": {
    "shCTL": [
      0,
      0
    ],
    "shKD": [
      0,
      0
    ],
    "delta": 0,
    "verdict": "DOES NOT MOVE AT THE LOCUS"
  },
  "ctcf_element": {
    "shCTL": [
      31.45,
      0
    ],
    "shKD": [
      20.15,
      17.7
    ],
    "delta": 3.2,
    "verdict": "DOES NOT MOVE AT THE LOCUS"
  },
  "ctcf_summit_1kb": {
    "shCTL": [
      32.51,
      0
    ],
    "shKD": [
      21.15,
      18.57
    ],
    "delta": 3.61,
    "verdict": "DOES NOT MOVE AT THE LOCUS"
  },
  "dcal_cut_1kb": {
    "shCTL": [
      32.54,
      87.72
    ],
    "shKD": [
      0,
      0
    ],
    "delta": -60.13,
    "verdict": "MOVES AT THE LOCUS - RULE FIRED, VERDICT NOT CREDIBLE, within-control spread is 55.18 points (deviation D2)"
  },
  "bcr_core": {
    "shCTL": [
      63.61,
      61.69
    ],
    "shKD": [
      56.56,
      0.09
    ],
    "delta": -34.32,
    "verdict": "DOES NOT MOVE AT THE LOCUS"
  },
  "kmt2a_tss_2kb": {
    "shCTL": [
      77.17,
      0
    ],
    "shKD": [
      71.52,
      59.19
    ],
    "delta": 26.77,
    "verdict": "DOES NOT MOVE AT THE LOCUS"
  },
  "kmt2a_gene": {
    "shCTL": [
      7.6,
      7.7
    ],
    "shKD": [
      2.56,
      2.64
    ],
    "delta": -5.05,
    "within_shCTL_spread": 0.1,
    "within_shKD_spread": 0.08,
    "verdict": "DOES NOT MOVE AT THE LOCUS - below the 10-point threshold, though it exceeds both replicate spreads fiftyfold"
  }
}
kmt2a gene against the matched gene null
{
  "shCTL": [
    18.18,
    20.26
  ],
  "shKD": [
    3.12,
    7.27
  ],
  "delta": -14.03,
  "note": "scored against N4, a null declared in section 6 for a DIFFERENT question; reported as secondary and not scored by the Q2 rule"
}
How this could bite us back
  • A KDM4 inhibitor would raise H3K9me3 at active-promoter +1 nucleosomes across the genome - 35.7 percentile points in this measurement - which is a transcription-wide effect, not a locus-directed one. Any proposal to use a KDM4 agent to restore H3K9 methylation at the KMT2A breakpoint would be buying a genome-wide pause-release perturbation to obtain, by this measurement, no change at the target.
  • The arm also shows the opposite risk for the KDM4C-targeting direction this graph already holds. gene-kdm4c records that KDM4C is required for MLL-fusion leukaemia and that SD70 preferentially kills MLL-fusion lines; this beat shows that removing KDM4B and KDM4C together does not move H3K9me3 at the KMT2A locus at all. Whatever SD70's therapeutic mechanism is in SEM, it is not a chromatin effect at this locus - or at least nothing here supports one.
Still open
  1. Does KDM4A maintain the deficit? The only public perturbation removes KDM4B and KDM4C and leaves KDM4A, the strongest H3K9me3 demethylase of the four, entirely intact. A KDM4A arm does not exist in any archive with an H3K9me3 readout.
  2. Is the −5.05 a loss or a failure to join in? A spike-in-normalised H3K9me3 ChIP under the same knockdown would separate the two, and no deposited experiment on this mark carries a spike-in.
  3. Why do the two knockdown replicates correlate with each other (rho 0.259) LESS well than a knockdown correlates with a control (0.290)? Q114 found the identical inversion in the +/-oestradiol comparison of this deposit's SUV39H1 tracks. Two independent occurrences in one deposit suggests a batch structure the sample metadata does not expose.
  4. Would the same knockdown move H3K9me3 at the KMT2A locus in a cell where the locus is NOT H3K9me3-poor to begin with? Every measurement here starts from a floor of 0.0000 at the element, and an eraser knockdown cannot raise a value that has no substrate.
Every source behind this entry (2)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Across the seven stored KMT2A intervals, the shKDM4B+shKDM4C arm differs from shCTL by less than the pre-registered 10-percentile-point threshold at the 90 kb gene body (Δ = −5.05, shCTL 7.60/7.70 and shKD 2.56/2.64, within-arm spreads 0.10 and 0.08) and does not move at the exon-11 element, which reads an explicit 0.0000 in all four tracks. Genome-wide over 56,171 anchors the median |Δpercentile| is 24.42 and only 9.85% of anchors move by more than their own within-arm replicate spread.

    What this evidence cannot show

    Two replicates per arm and no spike-in. HOMER makeUCSCfile normalises each bedGraph to a fixed tag total, so a uniform genome-wide change in the mark is invisible by construction; every statistic here is a percentile rank within each track's own distribution, which is invariant to that normalisation but also cannot see a global shift. The knockdown leaves KDM4A - UniProt's strongest H3K9me3 demethylase of the four - untouched, so this is a partial family knockdown. No knockdown-efficiency measurement is deposited, so the degree of KDM4B and KDM4C loss in these particular cells is unverified. All four tracks pass C1 (12.6x-13.4x over a printed non-zero denominator) and C1b (3.96x-4.65x over the chr19q13 KRAB-ZNF cluster).

    Evidence that argues against this (2)
    • THE PRE-REGISTERED ARM-INFORMATIVENESS TEST PASSES BY A HAIR AND HALF-FAILS. shCTL-vs-shCTL Spearman is 0.3247 against a mean across-condition 0.2898 - it passes. But shKD-vs-shKD is 0.2592, BELOW the across-condition value, so the two knockdown replicates agree with each other less than a knockdown agrees with a control. On the shKD side the failure signature Q114 found for the +/-E2 comparison is present. The arm is judged informative on the strength of the E2 positive control, not on this test. source
    • THE PRE-REGISTERED 'MOVES' RULE FIRED AT dcal_cut_1kb AND THE VERDICT IS NOT CREDIBLE. The two shCTL replicates read 32.54 and 87.72 there - a 55.18-percentile-point disagreement between two controls - while both shKD read 0.00, giving Δ = −60.13 which clears every clause of the rule. The rule lacks a ceiling on the within-control spread. Recorded as deviation D2; the threshold is not re-tuned and the verdict is printed as the rule returned it. source
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The double knockdown raises H3K9me3 at the +1 nucleosome (TSS+1 to TSS+200) of the 2,000 highest-expressed MCF-7 protein-coding genes from the 21.73rd to the 57.43rd percentile of a 4,000-window active-plus-silent set (Δ = +35.71), with no effect at the 1 kb promoter window (Δ = −4.79). The KMT2A +1 nucleosome moves from 0.00 to 35.25 (Δ = +35.25), indistinguishable from the genome-wide Δ.

    What this evidence cannot show

    Figure legend read via PMC8439151 on 2026-09-23; the figure itself was not inspected, so this is a reproduction of a DIRECTION stated in a legend, not of a plotted value. The gene sets and window geometry were declared in the script header before the tracks were reopened (deviation D7). The 200 bp windows are heavily zero-inflated - 13% to 81% of active-gene +1 nucleosome windows read exactly zero depending on the track - and the per-replicate active-percentile values are 43.45 and 0.00 (shCTL) against 71.69 and 43.18 (shKD), so THE WITHIN-ARM SPREAD IS OF THE SAME ORDER AS THE EFFECT and the two distributions overlap. The expression ranking comes from ENCODE MCF-7 RNA-seq (ENCFF251DYL), a different culture in a different laboratory from the ChIP.

    Evidence that argues against this (2)
    • With n = 2 per arm and per-replicate values of 43.45/0.00 against 71.69/43.18, the lower knockdown replicate (43.18) is indistinguishable from the higher control replicate (43.45). A two-sample comparison of that shape supports a direction and not an effect size, and a reader should not treat +35.71 as a measured magnitude. What makes it worth reporting is that the direction matches a published figure legend generated from these same samples by the people who made them. source
    • Because HOMER normalises to a fixed tag total, a genuine global rise in H3K9me3 appears as a rise at the responding regions AND a fall in percentile terms at the non-responding ones. The KMT2A gene body's −5.05 is therefore equally consistent with 'the locus loses the mark' and with 'the locus does not join a genome-wide rise'. The second is the more parsimonious reading given E2, and this entry adopts it. source
Connected to