Oncofusion

start / graph / genome-q114-the-kmt2a-exon-11-ctcf-summit-lands-within-12-bp-in-two-more-cell-types

Non-codingstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q114-suv39h1-at-the-locus

Two more independent cell types put the KMT2A exon-11 CTCF summit within 12 bp of the coordinate this graph carries: MCF-7 at chr11:118,488,689 (5 bp away) and SK-N-SH at chr11:118,488,682 (12 bp), both score-1000 optimal-IDR peaks - and in the same tracks the anchor ranks only 72.95th of 56,171 MCF-7 anchors, so it is an unambiguous anchor and an ordinary one

also called exon-11 CTCF summit in MCF-7 and SK-N-SH · the anchor is real and mid-rank, cell type 7

The exact spot inside the KMT2A gene that this project treats as a loop anchor shows up again, to within about ten letters of DNA, in a breast-cancer line and a nerve-cell line that nobody had checked. It is definitely a real anchor - and it is a middling one, not an exceptional one.

The same thing, for a scientist

**Two new independent confirmations of the coordinate, from peak sets fetched for an unrelated purpose** [E1][E2]. This graph carries the exon-11 CTCF summit at **GRCh38 chr11:118,488,694**, measured in HCT116 and confirmed by signal-weighted centroid in RPE within +/-34 bp (`genome-kmt2a-exon-11-ctcf-summit-measured`, `genome-kmt2a-exon-11-ctcf-summit-rpe-gray`). The two ENCODE CTCF peak sets used as this beat's class-matched null each carry a peak over the element: | cell | ENCODE file | peak (GRCh38, BED) | score | signalValue | summit offset (col 10) | **summit** | distance from 118,488,694 | |---|---|---|---:|---:|---:|---|---:| | **MCF-7** | ENCFF942TCG (ENCSR000AHD, optimal IDR) | chr11:118,488,624-118,488,727 | **1000** | 115.43 | 65 | **118,488,689** | **5 bp** | | **SK-N-SH** | ENCFF685KTA (ENCSR541AMF, optimal IDR) | chr11:118,488,426-118,488,989 | **1000** | 1048.85 | 256 | **118,488,682** | **12 bp** | Both carry the maximum narrowPeak score. **Four cell types now agree within 34 bp** - HCT116, RPE, MCF-7, SK-N-SH - across three labs and three assays' worth of processing. The coordinate is not in doubt. **And the same two files say the anchor is ordinary.** The MCF-7 CTCF signal track from the same experiment (ENCFF516HJR, `signal p-value`, GRCh38, `preferred_default`) places the 147 bp exon-11 window at the **72.95th percentile** of 56,171 MCF-7 CTCF summit windows - **mean 62.39** - which **fails** the 90th-percentile instrument control this beat pre-registered [E3]. `genome-kmt2a-exon-11-ctcf-anchor-is-a-mid-rank-ctcf-anchor-in-encode-cells` measured the same statistic in six other ENCODE cells: median **70.61**, range 28.15-81.76, **0 of 6 above the 90th**. **MCF-7 is the seventh cell type and 72.95 sits almost exactly on that median.** **So the finding replicates, in a cell type that entry did not cover, with a completely independent CTCF experiment and lab.** The pre-registered control keeps failing because its premise - that this is a top-decile CTCF anchor - is false, not because any pipeline is blind. The threshold was not moved; the arm was declared blind, which is the pre-registered consequence.

Go deeper
Why a 147 bp window centred elsewhere is not the reason

A tempting explanation for C4's repeated failure is geometry: the null windows are centred on each cell's own CTCF summits while the target is a fixed external 147 bp window from an RPE/Gray call, so any summit offset penalises the target.

The peak-level numbers above rule that out as the main effect. MCF-7's own summit is 5 bp from the centre of the fixed window and comfortably inside its 147 bp span; a 5 bp offset cannot move a percentile from the 90s to 72.95. The geometry worry is real in principle and should be fixed before C4 is reused - the honest form of the control centres on each cell's own summit - but here the element simply is a mid-rank anchor, which is what the six-cell entry already concluded.

What follows for every occupancy negative measured at this element

If the anchor were top-decile CTCF everywhere, a protein's absence from it would be surprising and informative. It is not: it is around the 70th percentile in seven of seven cell types tested. A factor-occupancy negative at a middling anchor is a weaker statement than the same negative at an exceptional one, and this applies to Q114's SUV39H1 negative, to Q112's BRCA1 and CBX3 reads, and to Q110's SETDB1 and CBX5 zeros alike.

It also means the class-matched null is the right null and the genome-wide one is badly wrong here: a mid-rank CTCF anchor is far above a random genome tile, so a genome-wide comparison would make any occupancy at this element look special when it is only as special as CTCF binding generally is.

Facts and figures
elementCTCF/cohesin loop-anchor site inside canonical KMT2A exon 11
interval grch38chr11:118488614-118488760
width bp147
graph stored summit grch38chr11:118488694
cell type measuredMCF-7 (ER+ breast adenocarcinoma) and SK-N-SH (neuroblastoma) - NEITHER haematopoietic
shown for fusionNO - CTCF occupancy at this element has never been measured in a KMT2A-rearranged cell of any lineage
motif orientationNOT MEASURED here either - still requires a PWM scan of the footprint sequence
new confirmations 2026 09 22
{
  "MCF-7": {
    "file": "ENCFF942TCG",
    "experiment": "ENCSR000AHD",
    "peak": "chr11:118488624-118488727",
    "peak_width_bp": 103,
    "score": 1000,
    "signalValue": 115.43089,
    "summit_offset": 65,
    "summit_grch38": 118488689,
    "bp_from_stored_summit": 5
  },
  "SK-N-SH": {
    "file": "ENCFF685KTA",
    "experiment": "ENCSR541AMF",
    "peak": "chr11:118488426-118488989",
    "peak_width_bp": 563,
    "score": 1000,
    "signalValue": 1048.84662,
    "summit_offset": 256,
    "summit_grch38": 118488682,
    "bp_from_stored_summit": 12
  }
}
cell types now agreeing within 34bp
[
  "HCT116",
  "RPE (signal-weighted centroid)",
  "MCF-7",
  "SK-N-SH"
]
c4 instrument control mcf7
{
  "file": "ENCFF516HJR",
  "output_type": "signal p-value",
  "exon11_mean": 62.39485224632368,
  "percentile_of_own_summit147_anchors": 72.95401541720817,
  "n_anchors": 56171,
  "threshold": 90,
  "passes": false
}
comparison to six encode cells q112
{
  "median_percentile": 70.61,
  "range": "28.15-81.76",
  "n_above_90th": 0,
  "n_cells": 6
}
How this could bite us back
  • A coordinate confirmed in four epithelial and neural lines still has no confirmation in the lineage the disease arises in, and the one blood measurement in the corpus (GM12878, 28.15th percentile) is the weakest of all of them. An intervention argument that depends on this being a strong anchor in a B-progenitor has no support.
Still open
  1. Does the C4 control pass anywhere if it is re-centred on each cell type's OWN summit rather than on the fixed RPE window? The 5 bp MCF-7 offset says the geometry is not the main effect here, but the control has now failed in seven of seven cell types and should be rebuilt correctly before an eighth beat inherits it.
  2. What is the CTCF motif orientation at this summit? Four cell types now agree on the position to within 12 bp and nobody has run a PWM scan of the 147 bp. Orientation decides which loop this anchor can form and is a one-command answer.
  3. Is the anchor mid-rank in a KMT2A-rearranged cell too? Every one of the seven cell types measured is a cell line with an intact 11q23. No CTCF experiment exists in any KMT2A-rearranged line in any archive, which is the absence absence-no-human-ctcf-or-cohesin-occupancy-measurement-in-a-blood-cell-at-kmt2a already records.
Every source behind this entry (3)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    ENCODE's MCF-7 CTCF optimal-IDR peak set contains a score-1000 peak at chr11:118,488,624-118,488,727 whose point-source summit is chr11:118,488,689, 5 bp from the summit this graph carries.

    What this evidence cannot show

    Read from the downloaded narrowPeak on 2026-09-22 (56,203 peaks, 32 on non-primary scaffolds). A narrowPeak point-source summit is a pipeline estimate, not a footprint; the +/-12 bp agreement across four cell types is the argument, not any single file's precision.

    Evidence that argues against this (1)
    • signalValue is not comparable across experiments: MCF-7 reads 115.43 and SK-N-SH 1048.85 at the same element, which reflects different depth and normalisation, not a nine-fold biological difference. Only the within-experiment percentile is interpretable. source
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    ENCODE's SK-N-SH CTCF optimal-IDR peak set contains a score-1000 peak at chr11:118,488,426-118,488,989 whose point-source summit is chr11:118,488,682, 12 bp from the summit this graph carries.

    ENCSR541AMF - CTCF ChIP-seq in SK-N-SH; optimal IDR thresholded peaks ENCFF685KTA (GRCh38) · Bradley Bernstein, Broad Institute · ENCODE Portal 2026
    What this evidence cannot show

    SK-N-SH is a neuroblastoma line and was chosen as the nearest available neural class-match for a diffuse midline glioma deposit, not for this purpose; the summit confirmation is a by-product. Its peak is 563 bp wide against MCF-7's 103 bp, so its summit estimate is the looser of the two.

    Evidence that argues against this (1)
    • CTCF site usage is substantially but not completely cell-type-invariant, so agreement across four lines does not establish that the site is occupied in a haematopoietic cell. Q112 measured it directly in GM12878 at the 28.15th percentile - present, and the weakest of the six. source
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    In the MCF-7 CTCF signal track from the same experiment, the exon-11 147 bp window sits at the 72.95th percentile of 56,171 MCF-7 summit-centred anchor windows, failing the pre-registered 90th-percentile instrument control - and matching the 70.61 median of six other ENCODE cell types.

    What this evidence cannot show

    The build of ENCFF516HJR was verified from its own chrom table (136 chroms; chr1 248,956,422, chr11 135,086,622, chrX 156,040,895 - GRCh38, hg19 excluded). The percentile is against that cell's own summit-centred windows, so it measures rank among CTCF anchors and nothing else.

    Evidence that argues against this (1)
    • A signal p-value track is not a fold-change track, and the six-cell comparison used fold-change over control. The two statistics rank similarly but are not identical, so the 72.95-versus-70.61 agreement is a close match between related rather than identical measurements. source
Connected to