Analysis (runnable): is anything actually enriched at KMT2A breakpoints? The test has never been run against a background that could say no
also called repeat enrichment test KMT2A breakpoints · Alu enrichment null model · breakpoint feature intersection
People have said for thirty years that the place KMT2A breaks is full of repeated DNA and other special features. Nobody has ever compared it to a fair stretch of DNA chosen the same way, so we do not know whether those features are enriched or simply present.
The same thing, for a scientist
The claim that the KMT2A breakpoint cluster region is repeat-rich rests on an indirect observation from 2003 - all four polymorphisms found on full resequencing of the 8.3 kb BCR in 87 individuals lay inside Alu elements [E1] - plus the annotation of Alu sequences at sequenced t(4;11) junctions in a paper whose purpose was PCR primer design [E2]. Neither is an enrichment test. Since then the breakpoint catalogue has grown to 3401 patients with determined genomic breakpoints [E3] and the gene's coordinate system has been re-standardised [E4], while repeat annotation, CTCF/cohesin maps, replication-timing maps and G-quadruplex predictions have all become routine public resources. The intersection has not been published: searches run 2026-09-20 return 0 records for KMT2A breakpoint with RepeatMasker, 0 with G-quadruplex, 0 with replication timing, 0 with CTCF at the breakpoint cluster region, and 0 with loop anchor or cohesin, all with clean querytranslations and no UID parsing artefact [E5]. This entry specifies the test and, more importantly, specifies the background, because the background is where this analysis will succeed or fail. The naive null - genome-wide repeat density - is wrong in a way that guarantees a positive result, since KMT2A breakpoints are confined to introns of one gene and intronic Alu density differs from the genomic average.
The background is the analysis1 deeper
Three nested nulls, each answering a different question, and a result should say which one it used:
Null A - within-gene. Permute breakpoint positions uniformly across the intronic sequence of KMT2A (GRCh38 chr11:118,436,456-118,526,832, 90,377 bp 6). This asks: given that a break happened somewhere in this gene, is it more likely to land in a repeat than chance? This is the only null that isolates a positional preference from the gene's own composition.
Null B - assay-matched. Permute only within regions the ascertaining assay could have detected. The recombinome breakpoints were obtained by long-distance inverse PCR and multiplex PCR with designed primer sets 3; a break in a region where the assay has no primer or where the amplicon is too long is invisible. Without this restriction, a repeat-density gradient across the gene will be confounded with a primer-coverage gradient, and the analysis will report the shape of the assay.
Null C - matched decoy genes. Compare against the intronic repeat composition of other large, highly transcribed haematopoietic genes that do NOT recurrently break. This asks whether KMT2A is unusual among genes of its class, which is the question anyone reading 'KMT2A is repeat-rich' actually thinks is being answered.
Running only Null A and reporting a p-value would be the commonest version of this analysis and the least informative.
Why this is not the same analysis as the linker decomposition
ana-breakpoint-linker-vs-fragility-decomposition regresses breakpoint counts on position with repeat density among its covariates, and asks about a protein-structural coefficient after adjustment. That analysis assumes the fragility covariates are meaningful and estimates something else. This one asks whether the fragility covariates have any enrichment signal at all, and under which null. If this analysis returns no repeat enrichment under Nulls A and B, the covariate in the other analysis is adjusting for noise and its residual coefficient means something different. They should be run in this order.
What is worth intersecting, ranked by how cheap it is
- RepeatMasker classes - Alu/SINE, L1/LINE, MIR, simple repeats, low-complexity - from the UCSC repeat track on GRCh38 7. Hours of work.
- CTCF and cohesin occupancy plus called loop anchors in a haematopoietic progenitor. This is the single highest-value intersection, because the loop-anchor fragility model 8 makes a concrete prediction about the BCR that this graph's
breakpoint-kmt2a-bcr-architectureexplicitly asks about, and the answer does not exist in the literature:(MLL OR KMT2A) AND (loop anchor OR cohesin) AND breakpointreturns 0 5. - Replication timing and initiation-zone maps. Also 0 records at this locus 5.
- Predicted G-quadruplexes and non-B DNA. Also 0 records 5. Note this graph already flags that no measurement of supercoiling, torsional stress or R-loop density exists across the KMT2A BCR.
- ENCODE candidate cis-regulatory elements, to test whether the DNase-hypersensitive cryptic promoter inside the BCR is the dominant feature rather than any repeat.
The power problem, stated honestly
The major BCR is 8.3 kb. Even with thousands of breakpoints, an 8.3 kb window contains only a handful of independent repeat elements, so a within-BCR enrichment test has very few effective degrees of freedom whatever the sample size - the breakpoints are pseudo-replicates of a fixed annotation. The test only becomes well-powered across the whole 90 kb gene and across the second BCR and the partner loci, where the annotation varies. An enrichment analysis restricted to the 8.3 kb BCR will be underpowered no matter how many patients are in it, and a null result there should not be reported as evidence of no enrichment.
The matching problem on the partner side is worse still: AFF1 spans 206,871 bp on GRCh38 6 and its breakpoint region has never been characterised structurally to the standard KMT2A's has, which this graph records as an open question.
| script | analyses/kmt2a-breakpoint-feature-enrichment/run.py |
|---|---|
| status | planned |
| power limitation | an enrichment test confined to the 8.3 kb major BCR has very few independent annotation elements and is underpowered regardless of patient number; a null result there is uninformative |
| cost tier | low |
| result summary | not run |
inputs
[ "KMT2A breakpoint positions re-mapped to GRCh38 using the 2024 consensus gene structure", "UCSC RepeatMasker track, GRCh38, library version recorded", "CTCF and cohesin (RAD21) ChIP-seq plus called loop anchors in a haematopoietic progenitor", "Repli-seq or initiation-zone map in a haematopoietic progenitor", "predicted G-quadruplex and non-B DNA tracks", "ENCODE candidate cis-regulatory elements", "primer/amplicon coverage of the ascertaining assay, for the assay-matched null" ]
nulls
[ "A: uniform permutation over KMT2A intronic sequence, GRCh38 chr11:118436456-118526832", "B: permutation restricted to regions the ascertaining PCR assay could detect", "C: matched decoy genes - large, highly transcribed haematopoietic genes that do not recurrently break" ]
gene spans grch38
{
"KMT2A": "chr11:118436456-118526832 (+), 90377 bp",
"AFF1": "chr4:86934203-87141073 (+), 206871 bp",
"source": "Ensembl REST, retrieved 2026-09-20"
}counted absences 2026 09 20 — the numbers behind the chart
[
{
"q": "(MLL OR KMT2A) AND breakpoint AND RepeatMasker",
"count": 0
},
{
"q": "(MLL OR KMT2A) AND breakpoint AND (G-quadruplex OR G4)",
"count": 0
},
{
"q": "(MLL OR KMT2A) AND breakpoint AND (replication timing)",
"count": 0
},
{
"q": "(MLL OR KMT2A) AND breakpoint cluster region AND CTCF",
"count": 0
},
{
"q": "(MLL OR KMT2A) AND (loop anchor OR cohesin) AND breakpoint",
"count": 0
},
{
"q": "(MLL OR KMT2A) AND breakpoint AND (permutation OR background model OR enrichment analysis)",
"count": 6
}
]- A positive repeat enrichment would be descriptive, not actionable: nobody can remove Alu elements from a fetal genome, and the finding would only redirect which mechanism to study.
- A cell-type-independent architectural explanation for where the break lands would explain the position while explaining nothing about why this disease occurs in one narrow progenitor window, which is the question the PI actually asked.
- Does the KMT2A major BCR, and specifically the intron 11 / exon 12 hotspot, overlap a called CTCF/cohesin loop anchor in a haematopoietic progenitor? This is the highest-value single intersection and it returns 0 published records.
- Is Alu density inside the KMT2A BCR higher than in matched introns of the same gene, or than in decoy genes of similar size and expression? No published comparison exists in either direction.
- Can the assay-coverage map of the ascertaining PCR panels be obtained? Without it, Null B cannot be built and every positional claim in this literature stays confounded with primer design.
- Does the breakpoint density track replication timing? Common fragile sites are late-replicating, and if KMT2A is not, the transcription-replication conflict route this graph carries elsewhere loses one of its cheapest predictions.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Full resequencing of the 8.3 kb KMT2A BCR in 87 individuals found four polymorphic sites and every one lay inside an Alu repetitive element, while the three poly-A tract polymorphisms occurred at equal frequency in leukaemia patients and controls.
Polymorphisms in the MLL breakpoint cluster region (BCR) · Echlin-Bell DR, Smith LL, Li L, et al. · Human genetics 2003What this evidence cannot show
This is an indirect argument for repeat density and the paper was not designed as an enrichment study. Four polymorphisms is a small denominator. The study's own risk finding is negative, which is the part most often dropped when it is cited.
Evidence that argues against this (1)
- Alu elements are the most abundant repeat family in the human genome and are strongly enriched in introns generally, so finding four polymorphisms inside Alu in an intronic region is close to what an unremarkable stretch of intron would give. The observation constrains the region's composition; it does not distinguish KMT2A from background, and no published analysis has made that comparison at this locus. Q32-break-machinery 2026
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Alu elements are annotated at and around sequenced t(4;11) genomic breakpoints in the KMT2A and AFF1 breakpoint cluster regions.
Breakpoints of t(4;11) translocations in the human MLL and AF4 genes in ALL patients are preferentially clustered outside of high-affinity matrix attachment regions · Hensel JP, Gillert E, Fey GH, Marschalek R · Journal of cellular biochemistry 2001What this evidence cannot show
The paper's subject is matrix attachment, not repeat enrichment, and its headline finding is that breakpoints cluster OUTSIDE high-affinity matrix attachment regions. Presence of a feature at a junction is not enrichment of that feature at junctions.
Evidence that argues against this (1)
- This paper's own conclusion is the opposite in sign to the other primary mapping study of the same region, which placed a high-affinity scaffold attachment region inside the telomeric half of the BCR alongside six TOP2 consensus sites and found therapy-related breakpoints concentrated there. The two are not reconciled, so any feature-association claim about the BCR built on either one inherits an unresolved disagreement - which is itself the strongest argument for running the enrichment test with an explicit null rather than citing a paper. Broeker PL et al. 1996
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Genomic breakpoints within KMT2A and its translocation partner genes have been determined for 3401 acute leukaemia patients, with the seven most frequent partners plus partial tandem duplication accounting for more than 90% of all recombinations.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
One referral-based diagnostic centre over 20 years, so the case mix is not a population sample. Breakpoints were ascertained by PCR-based methods with designed primer sets, which is precisely the ascertainment bias Null B is designed to absorb.
Evidence that argues against this (1)
- An absence of detection is not an absence of breaks. A targeted PCR panel returns negative for every partner and every position it was not designed around, so a breakpoint density map built from such an assay measures the union of biology and primer design. The graph's standing rule that 'an absence of testing is not a negative result' applies directly, and it is why the enrichment null must be assay-matched rather than sequence-matched. Meyer C et al. 2024
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
A consensus KMT2A/MLL gene structure has been published as a comprehensive update for research and diagnostic use, standardising exon and intron numbering.
The KMT2A/MLL consensus gene structure: a comprehensive update for research and diagnostic implications · Meyer C, Larghero P, Lopes BA, Marschalek R · Leukemia 2024What this evidence cannot show
Read from the record and title; cited here only for the existence of a standardised structure, which any coordinate-based intersection needs. This entry does not assert any specific exon numbering from it.
Evidence that argues against this (1)
- The existence of a 2024 consensus is itself evidence that the pre-2024 literature is on inconsistent coordinates. Every breakpoint position published before it must be re-mapped rather than trusted, and any repeat-intersection assembled from historical papers without re-mapping will mis-assign positions by whole introns. Hasan SK et al. 2014
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database
PubMed searches run 2026-09-20 return zero records linking KMT2A breakpoints to RepeatMasker annotation, G-quadruplexes, replication timing, CTCF at the breakpoint cluster region, or loop anchors and cohesin, in each case with a clean querytranslation containing no UID parsing artefact.
PubMed E-utilities counted-absence searches for genome-feature intersections at the KMT2A breakpoint region, run 2026-09-20; five queries, all count 0, querytranslation inspected for each · Q32-break-machinery · NCBI E-utilities 2026What this evidence cannot show
Title/abstract indexing only, and the symbol set is MLL and KMT2A. An intersection reported as a supplementary panel of a paper about something else would not appear. Each querytranslation was read to exclude the parenthesis/UID failure mode the Queen brief documents; none showed a [UID] term.
Evidence that argues against this (1)
- One of these zeros is a near-miss rather than a true absence. Genome-wide, TOP2-mediated DSBs at CTCF/cohesin loop anchors have been mapped and the genes that recurrently translocate to drive leukaemias were reported as enriched there, with MLL fusions specifically studied. The locus-level intersection is missing; the class-level claim is not. Writing 'nobody has looked at CTCF and KMT2A breakage' without that qualification would be wrong. Gothe HJ et al. 2019
- E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
On GRCh38, KMT2A spans chr11:118,436,456-118,526,832 on the plus strand (90,377 bp) and AFF1 spans chr4:86,934,203-87,141,073 on the plus strand (206,871 bp).
Ensembl REST lookup, genes ENSG00000118058 (KMT2A) and ENSG00000172493 (AFF1), assembly GRCh38, retrieved 2026-09-20 · Ensembl · Ensembl REST API 2026What this evidence cannot show
Canonical-gene span on the primary assembly. Alternative transcripts and the historic BamHI-fragment coordinate system used throughout the 1990s breakpoint literature do not map onto these numbers without explicit lift-over.
Evidence that argues against this (1)
- A database span is a scope-limited fact in exactly the way the Queen brief warns about: it gives the extent of the annotated gene, not the extent of the region in which breaks occur. KMT2A has at least two breakpoint cluster regions plus a documented second BCR, so 'within the gene' and 'within the breakpoint-competent region' are different intervals, and a permutation over the gene span will over-represent intervals that never break. Meyer C et al. 2023
- E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
Repeat annotation for GRCh38, including Alu/SINE, LINE, MIR, simple-repeat and low-complexity classes, is available as a public track that can be intersected with arbitrary coordinate sets.
UCSC Genome Browser RepeatMasker track, GRCh38/hg38 · UCSC Genome Browser · UCSC Genome Browser 2026What this evidence cannot show
RepeatMasker output is a computational annotation with a library version; results differ between library releases, so the analysis must record the library and assembly used. Repeat calls in low-mappability regions are themselves less reliable.
Evidence that argues against this (1)
- A repeat annotation existing is not the same as the enrichment question being well posed. The standing database-scope caution applies: an annotation track records what an alignment program found against a library, which is a statement about sequence similarity and not about the biophysical property - recombinogenicity, torsional stress, secondary structure - that a breakage claim actually needs. Echlin-Bell DR et al. 2003
- E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
CTCF/cohesin-bound chromosome loop anchors accumulate TOP2B-mediated double-strand breaks and DSBs frequently occur at breakpoint clusters commonly translocated in cancer.
Genome Organization Drives Chromosome Fragility · Canela A, Maman Y, Jung S, et al. · Cell 2017What this evidence cannot show
Genome-wide claim; no measurement at the KMT2A BCR. Cited here to justify the CTCF/cohesin intersection as the highest-value one to run, not as evidence that the intersection will be positive.
Evidence that argues against this (1)
- Loop-anchor breakage is reported in this paper as largely cell-type-independent, which is awkward for using it to explain a leukaemia that arises in one narrow progenitor compartment. If the intersection turns out positive at the KMT2A BCR in every cell type examined, it will have explained the position of the break and explained nothing about the cell in which it matters. Canela A et al. 2019
the modern repeat-annotation intersection that entry asks for, with the null model specified
asks whether the KMT2A BCR overlaps a called CTCF/cohesin loop anchor in a haematopoietic progenitor
re-tests the SAR/TOP2/repeat feature claims of the 1996-2003 literature against an explicit background
must run first: that analysis adjusts for fragility covariates this one asks whether to believe
- related to →KMT2A / MLL1 — the gene that breaks
GRCh38 span chr11:118436456-118526832, 90377 bp, used as the permutation domain
includes the G-quadruplex/non-B-DNA intersection that hypothesis needs and that returns 0 published records
supplies the one published non-translocating comparator gene, the control that analysis generalises