Oncofusion

start / graph / ana-q116-kap1-has-a-called-peak-on-the-kmt2a-promoter-in-mcf7-and-none-within-52-kb-of-the-anchor

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

RUN: KAP1/TRIM28 has a called peak overlapping the KMT2A promoter in MCF-7 in both hormone conditions, 267 bp of it inside the K562 TRIM28 peak Q110 found - and its nearest chr11 peak to the exon-11 CTCF summit is 52,230 bp away

also called the TRIM28 promoter address, replicated · KAP1 peak level at the KMT2A locus · GSE139196 does deposit peak files

A scaffolding protein that normally helps silence DNA sits right on the KMT2A gene's on-switch in breast cells, at almost exactly the same spot it was already found in blood cells. It is nowhere near the place where the gene breaks.

The same thing, for a scientist

**The result.** HOMER `findPeaks -style factor` output deposited with GSE139196's two 2021 KAP1 samples calls a peak at **chr11:118,436,186-118,436,464** in MCF-7 without oestradiol and **chr11:118,436,251-118,436,545** with it - straddling the *KMT2A* canonical TSS region. It is the **13th and 6th strongest peak on chr11** in the two files, and the 1,512th of 8,278 and 1,062nd of 5,884 genome-wide (top 18.3% and 18.1%) [E1]. **And it replicates an address this graph already holds, across two lineages.** `ana-q110-h3k9-writer-and-reader-occupancy-at-the-kmt2a-locus` found a TRIM28 ENCODE conservative-IDR peak at **chr11:118,436,029-118,436,453** in **K562**. The MCF-7 -E2 peak overlaps it by **267 bp** and the +E2 peak by **202 bp**. Two cell lineages (erythroleukaemia and breast), two labs, two peak callers, two peak-set conventions - **same 270 bp of the *KMT2A* promoter** [E1][E2]. **Nothing at the breakpoint.** Zero called peaks in the 146 bp exon-11 window, zero in the 7,465 bp BCR core, zero in the 1,047 bp CTCF element, in both conditions. The **nearest chr11 peak to the exon-11 CTCF summit (118,488,694) is 52,230 bp away in -E2 and 52,149 bp in +E2**. This matches Q110's K562 peak-level zero exactly [E1]. **A correction this analysis had to make first.** `analyses/q114-suv39h1-at-the-kmt2a-locus/PREREGISTRATION.md` records that GSE139196 deposits **no peak file** ("HOMER `findPeaks` was run but no peak file is in the deposit"). **For GSM5660804 and GSM5660805 it does** - each carries a `*.hg38.txt.gz` of 184,951 and 134,236 bytes that the sample record itself calls a peak file (`Supplementary_files_format_and_content: bedgraph and peak file`). Those two files are the whole peak-level arm of this entry, and the corpus would not have known they existed [E1]. **AND THE POSITIVE CONTROL FAILED, which bounds the negative and not the positive.** The *ZNF274* gene body (chr19:58,183,029-58,215,964) - the canonical ZNF274-TRIM28-SETDB1-ATRX target class, where Q110's ENCODE TRIM28 set has 2 peaks - carries **0 peaks in both files**, and the chr19q13 KRAB-ZNF 1 Mb cluster carries **2 and 1 against 2.67 and 1.90 expected under uniform placement**, i.e. exactly the same as the chr2 gene desert. **A peak-level absence from these two files is therefore weak**, and I read the BCR zero as consistent with Q110 rather than as independent confirmation. The promoter peak is unaffected: a called peak is a called peak.

Go deeper
Every interval, both files

Peaks overlapping each query, with the count expected if the file's peaks were placed uniformly over 3.1 Gb:

intervalqueryGSM5660804 (-E2)GSM5660805 (+E2)expected (-E2)
exon11 (146 bp)chr11:118,488,614-118,488,760000.0004
ctcf_elementchr11:118,488,159-118,489,206000.003
bcr_corechr11:118,482,422-118,489,887000.020
dcal_cut_1kbchr11:118,484,362-118,485,361000.003
kmt2a_tss_2kbchr11:118,435,456-118,437,456110.005
kmt2a_genechr11:118,436,456-118,526,832110.241
kmt2a_gene +/-50 kbchr11:118,386,456-118,576,832210.508
Q110's K562 TRIM28 promoter peakchr11:118,436,029-118,436,453110.001
Q110's K562 shared upstream elementchr11:118,415,740-118,416,164000.001
ZNF274 gene body (POSITIVE CONTROL)chr19:58,183,029-58,215,9640 - FAILS0 - FAILS0.088
chr19q13 ZNF cluster 1 Mbchr19:52,000,000-53,000,000212.670
chr2 gene desert (NEGATIVE)chr2:35,000,000-36,000,000212.670

The kmt2a_tss_2kb hit is 1 peak against 0.005 expected - a 200-fold excess over uniform placement, and it is the same peak in both conditions. The q110_k562_shared_upstream_element zero is a real difference from K562: the second H3K9-machinery address Q110 found 20 kb upstream of KMT2A does not reproduce in MCF-7, although a different KAP1 peak sits at chr11:118,401,606-118,401,884, 14 kb further away.

Why the positive control failed, and what that does and does not cost

Both files were called with findPeaks -center -style factor, with fold over local region required = 4.00 over a 10,000 bp local window and the putative peak list cut from 72,376 to 8,278 by that local filter alone. A point-source caller with a 4x local-fold requirement cannot call a broad domain, because a domain is not enriched over its own 10 kb neighbourhood. KAP1 at the 3' zinc-finger-coding exons of KRAB-ZNF genes is a domain-shaped signal. So the ZNF274 zero is the expected behaviour of this caller on this target class rather than evidence the ChIP failed - but that is an explanation I am offering, not something these files measure, and the pre-registered control is recorded as FAILED and is not re-tuned.

The consequence is asymmetric and worth stating plainly: the BCR and exon-11 zeros from these two files carry little weight on their own, and are quoted here only because they agree with Q110's ENCODE conservative-IDR zero in K562, which had a passing positive control. The promoter peak is unaffected by the control failure in either direction.

Facts and figures
scriptanalyses/q116-h3k9-axis-at-the-locus/kap1_peaks.py
prereganalyses/q116-h3k9-axis-at-the-locus/PREREGISTRATION.md
statusrun
run date2026-09-23
run byQ116-h3k9-axis-at-the-locus
complete runtrue
assemblyhg38 / GRCh38 native, stated by the sample records and by the file names; nothing lifted
peak callerHOMER findPeaks -center -style factor, FDR 0.001, peak size 279 bp, fold over local region 4.00 over a 10 kb window, 72,376 putative peaks filtered to 8,278 by the local filter
positive control ZNF274 gene bodyFAILED - 0 peaks in both files, against 2 in Q110's ENCODE K562 TRIM28 set. Recorded as failed and not re-tuned. Consequence: the peak-level absences from these files are weakly bounded and are quoted only because they agree with a peak set whose control passed.
negative control chr2 desert2 and 1 peaks in a 1 Mb window, identical to the chr19q13 KRAB-ZNF cluster in the same files
correction to an existing analysis fileanalyses/q114-suv39h1-at-the-kmt2a-locus/PREREGISTRATION.md section 2 states that GSE139196 deposits no peak file. GSM5660804 and GSM5660805 each deposit one. Verified from the sample records and by download on 2026-09-23.
localizationchromatin at chr11q23.3 - the KMT2A promoter, the gene body, the 7,465 bp BCR core and the 146 bp exon-11 CTCF anchor, plus 50 kb flanks; chr19q13 (ZNF274 and the KRAB-ZNF cluster) and a chr2 gene desert as controls
subnuclear localizationnot measured
cell typeMCF-7 (ER+ breast adenocarcinoma). NOT haematopoietic, NOT KMT2A-rearranged.
cell statehormone-deprived culture, and 40 min 17-beta-oestradiol
epigenetic readoutoccupancy of KAP1/TRIM28, the H3K9me3 corepressor scaffold and HP1 partner, at the KMT2A locus
shown for fusionNO - neither this cell line nor K562 carries a KMT2A fusion, and nobody has measured TRIM28 occupancy at this locus in a KMT2A-rearranged cell of any lineage
absence classmeasured-not-analysed
absence class noteBoth peak files were deposited publicly and nobody had opened them; Q114's pre-registration recorded, incorrectly, that they do not exist.
results
[
  "analyses/q116-h3k9-axis-at-the-locus/results_kap1_peaks.json"
]
gsm5660804 minusE2
{
  "n_peaks": 8278,
  "n_peaks_chr11": 295,
  "total_tags": 61264298,
  "approximate_ip_efficiency": "0.77%",
  "kmt2a_promoter_peak": "chr11:118,436,186-118,436,464",
  "peak_id": "chr11-13",
  "normalized_tag_count": 11.8,
  "findpeaks_score": 75,
  "rank_genome_wide": 1512,
  "rank_percentile_genome_wide": "top 18.27%",
  "overlap_with_k562_trim28_peak_bp": 267,
  "distance_nearest_chr11_peak_to_exon11_summit_bp": 52230
}
gsm5660805 plusE2
{
  "n_peaks": 5884,
  "n_peaks_chr11": 235,
  "total_tags": 65374891,
  "approximate_ip_efficiency": "0.56%",
  "kmt2a_promoter_peak": "chr11:118,436,251-118,436,545",
  "peak_id": "chr11-6",
  "normalized_tag_count": 12.2,
  "findpeaks_score": 81,
  "rank_genome_wide": 1062,
  "rank_percentile_genome_wide": "top 18.05%",
  "overlap_with_k562_trim28_peak_bp": 202,
  "distance_nearest_chr11_peak_to_exon11_summit_bp": 52149
}
signal level companion added 2026 09 23
{
  "note": "the four 2019 KAP1 (TRIM28) ChIP-seq bedGraphs were read at signal level in the same run",
  "kmt2a_promoter_pct_of_2000_ACTIVE_promoters": {
    "GSM4133828_minusE2": 98.9,
    "GSM4133829_minusE2_rep2": 98.8,
    "GSM4133830_plusE2": 99.35
  },
  "x_median_active_promoter": [
    3.02,
    3.85,
    2.98
  ],
  "exon11_pct_of_ctcf_anchors": {
    "GSM4133828": 90.62,
    "GSM4133830": 0
  },
  "reading": "the peak-level promoter call and the signal-level 98.8th-99.35th percentile of active promoters are the same observation by two instruments; the exon-11 signal numbers span 90.62 to 0.00 across two tracks and support nothing"
}
How this could bite us back
  • TRIM28 is the scaffold through which SETDB1 silences endogenous retroelements and through which ATM signals at double-strand breaks via KAP1-S824 phosphorylation. If the KMT2A promoter is a constitutive KAP1 address across lineages, then any agent that disables KAP1 - or the ATM inhibitors already in trials - changes something at this promoter, 52 kb from a breakpoint that this project argues is break-prone. Nothing here measures that consequence and the direction is not predictable from occupancy.
  • This is an occupancy fact with no function attached. A peak at a promoter in two cancer lines neither predicts nor excludes an effect on KMT2A transcription, and this graph already records that TRIM28's link to the fusion itself is p = 0.129 on 3 and 2 spectra. Treating a promoter peak as a therapeutic handle here would be building on the weakest link in the chain.
Still open
  1. Do KAP1 and G9a co-occupy the KMT2A promoter in MCF-7 as they do in K562? This run places both there - KAP1 at peak level, G9a at the 95.8th signal percentile - but never intersected their genome-wide sets, so the base rate for co-occupancy is unknown and the pairing is a coincidence with no denominator. It is one bedtools call and Q110 asked the same question of K562 and also did not run it.
  2. Are the two 2021 KAP1 samples the same antibody as the four 2019 KAP1 (TRIM28) samples? The deposit states no clone for either and the titles differ ('ChIP-seq on KAP1' versus 'ChIP on KAP1 (TRIM28)'). Q114 raised this and it remains unanswerable from GEO.
  3. Does the ZNF274 control fail because HOMER's local filter cannot call a domain, or because these two ChIPs did not work? Re-calling the deposited 636 MB bedGraphs at domain scale would separate the two and this run could not afford the download.
  4. Is the KMT2A promoter a KAP1 address in a haematopoietic cell that is NOT K562? Two lineages agree, and one of them is the leukaemia line this project keeps substituting. A CD34+ or lymphoid TRIM28 peak set would make this a general property of the locus rather than a coincidence of two cancer lines.
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    GSM5660804 and GSM5660805 each deposit a HOMER findPeaks factor-style peak file (184,951 and 134,236 bytes, 8,278 and 5,884 peaks, hg38). Both call one peak overlapping chr11:118,435,456-118,437,456 (the KMT2A TSS +/-1 kb window) - chr11:118,436,186-118,436,464 and chr11:118,436,251-118,436,545 - and zero peaks in the exon-11 window, the CTCF element or the BCR core. The nearest chr11 peak to the exon-11 CTCF summit at 118,488,694 is 52,230 and 52,149 bp away. The ZNF274 gene-body positive control returns zero peaks in both.

    GSE139196: ChIP-seq on KAP1 in MCF-7 with and without estradiol, GSM5660804 and GSM5660805, hg38 HOMER peak files · Yang F, Tanasa B, Micheletti R, Ohgi KA, Aggarwal AK, Rosenfeld MG · NCBI GEO 2021
    What this evidence cannot show

    One sample per condition, no replicate, no IgG, no input-subtracted peak set, and no antibody clone stated. The declared positive control (ZNF274 gene body) FAILED in both files and the chr19q13 KRAB-ZNF cluster reads the same peak density as a chr2 gene desert, so this peak set's ABSENCES are weakly bounded; only its calls carry weight. The two companion bedGraphs (636 MB and 684 MB) were out of this run's disk budget, so no signal-level number from these two samples exists here - the signal-level KAP1 numbers in this beat come from the four 2019 GSM4133828-31 tracks instead. Content-Length verified for both peak files on 2026-09-23 before download and asserted again after.

    Evidence that argues against this (1)
    • The promoter peak is 11.8 and 12.2 normalized tags against a top peak of 1,525.5 in the same file - the 1,512th of 8,278 and 1,062nd of 5,884 genome-wide. It is a real call in the top fifth of the file, not a strong one, and a reader should not read 'KAP1 is at the KMT2A promoter' as 'KAP1 is concentrated there'. What makes it worth recording is the 200-fold excess over uniform placement in a 2 kb window and the 267 bp overlap with an independent call in a different lineage, not the peak's own height. source
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    Q110 found a TRIM28 conservative-IDR ENCODE peak at chr11:118,436,029-118,436,453 in K562 (signalValue 51.46 at the promoter, from ENCSR474CVP/ENCFF382UDO, 12,075 genome-wide peaks) and zero TRIM28 peaks in the KMT2A gene body, the BCR or the CTCF element.

    What this evidence cannot show

    K562 is a CML-derived erythroleukaemia line, not KMT2A-rearranged; one TRIM28 peak set. That entry's own positive control (TRIM28 peaks in the ZNF274 gene body) PASSED with 2 peaks, which is exactly the control that fails in the MCF-7 HOMER files here - so the two peak sets are not equally bounded and only the shared positive call is a true replication.

    Evidence that argues against this (1)
    • The two peak sets also DISAGREE about the second address. Q110 reports EHMT2 and TRIM28 both at chr11:118,415,7xx-118,416,1xx, an element about 20 kb upstream of the gene; the MCF-7 KAP1 peak files call zero peaks there. Either that element is K562-specific or the HOMER local filter removed it. This analysis cannot distinguish the two and does not claim the upstream element replicates. source
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    KAP1 is TRIM28 (Q13263, 835 aa), a major HP1-interacting partner identified by mass spectrometry in this same MCF-7 system, and its recruitment to oestrogen-induced enhancers and transfer to cognate promoters is required for E2-induced Pol II pause release.

    Shape of promoter antisense RNAs regulates ligand-induced transcription activation · Yang F, Tanasa B, Micheletti R, Ohgi KA, Aggarwal AK, Rosenfeld MG · Nature 2021
    PMID 34194047DOI
    What this evidence cannot show

    Full text read via PMC8439151 on 2026-09-23. The paper's KAP1 work is at ER-responsive enhancers and promoters; it never mentions KMT2A (zero hits for 'KMT2A' and for 'MLL' in the full text). The identity KAP1 = TRIM28 matters because UniProt's entry name KAP1_HUMAN is P31321 / PRKAR1B, a 381 aa kinase regulatory subunit - a trap already recorded in data/symbol-collisions.json and checked before this entry was written.

    Evidence that argues against this (1)
    • This graph's own gene-trim28 entry records that TRIM28's association with KMT2A::AFF1 is p = 0.129 with 3 and 2 spectra, i.e. possibly noise, and asks whether TRIM28 associates with the fusion at all when measured by a method that is not rank-truncated. A promoter occupancy fact in two non-leukaemic lines does not answer that, and should not be read as evidence that TRIM28 is in the fusion's orbit. source
Connected to