The experiment: cross-tabulate MLLT1 breakpoint class against age, lineage and outcome in the Frankfurt archive, then confirm reciprocal status by RNA in whoever has it
also called MLLT1 control-arm study protocol · 153 vs 101 study
Three steps, all on data that already exists. First check that the two patient groups are alike in age and disease type; then link them to how they did; then confirm on a subset that the mirror-image gene really is absent in one group and present in the other.
The same thing, for a scientist
A three-stage, no-new-sample design on the DCAL Frankfurt archive that supplied both the 460-case recombinome row and the 302-case t(11;19) breakpoint distribution [E1][E2]. **Stage 1 (balance check, days):** cross-tabulate MLLT1 breakpoint class - intergenic ACER1-MLLT1 versus MLLT1 intron 1 versus downstream-of-exon-2 - against age group and against lineage. This is the go/no-go. The only published test is on the 254-versus-48 YEATS split (age p = 0.33; lineage p < 0.0001) and says nothing about the 153-versus-101 split [E2]. If age or lineage is unbalanced, the comparison is confounded and stages 2 and 3 should not be run as an outcome study. **Stage 2 (outcome linkage, months):** for the 153 and 101, recover event-free survival, overall survival and end-of-induction MRD from the referring national study groups. **Stage 3 (label validation, weeks):** in the subset with archived RNA, assay the reciprocal junction directly by RT-PCR or by counting reciprocal junction reads in existing RNA-seq, to measure the leak in both arms - reciprocal-positive patients misassigned to the negative arm via breaks inside ACER1 [E3], and reciprocal-negative patients misassigned to the positive arm via 3'-KMT2A deletion or out-of-frame junctions [E4].
Why stage 1 comes first and is a genuine go/no-go
KMT2A-side breakpoint preference in ALL changes with age at about six months 1, MLLT1 is 52.6% infant 1, and lineage flips this partner's prognostic sign 5. If the intergenic-break class is infant-enriched, a survival difference between the arms is an age effect and the study should not be reported as a reciprocal-status result. This stage costs one query against a database that already holds breakpoint class, age and disease for every patient, and it converts the whole proposal from speculative to either powered or dead. Running stages 2 and 3 before stage 1 is the failure mode to avoid, because an unbalanced comparison that produces a positive result is much harder to withdraw than one that was never run.
Controls, and what each rules out
- Negative control arm: the 48 downstream-of-exon-2 patients. They differ from the 153 in direct-product architecture and in reciprocal capability. If they behave like the 101, reciprocal status is doing the work; if they behave like neither, architecture is.
- Partner control: run the identical breakpoint-class analysis in EPS15 (12 spliced of 51) and ELL (10 of 119) 1. Both are far too small to power an outcome comparison, but if the direction replicates in two independent partners the MLLT1 result is much harder to dismiss.
- Null-partner control: AFF1, where the spliced class is 1 of 1,262 1. Any analysis pipeline that finds a breakpoint-class effect in AFF1 has a bug.
- Assay control for stage 3: a parallel assay for the direct junction in every sample, so that a failed reciprocal RT-PCR can be distinguished from a failed RNA preparation. Without it a negative is uninterpretable, which is the standing lesson from the phospho-contacts beat.
Power, stated honestly
With arms of 153 and 101 and a binary endpoint near 50%, the design has roughly 80% power to detect an absolute difference of about 19 percentage points at two-sided alpha 0.05 (this graph's calculation). For a time-to-event endpoint the achievable precision depends on event count, which is unknown until stage 2. This study can detect a large effect and cannot exclude a moderate one. A null result is a bounded null - 'no difference of 19 points or more' - and must be written that way. Given that the t(4;11) reciprocal-status association was reported as an EFS gap of 62.4% versus 11.7%, an effect of the size previously claimed would be detected easily; an effect half that size would not be excluded.
The cheaper computational variant, if the archive is inaccessible
Reciprocal status may be callable retrospectively from existing RNA-seq: a reciprocal MLLT1::KMT2A junction produces chimeric reads that standard fusion callers discard as the 'reciprocal artefact' of the direct call. Re-examining TARGET, St. Jude Cloud or the KMT2A-r cohorts already deposited in GEO for reciprocal junction reads, in samples with known outcome, would give an independent and fully public version of stage 2 and stage 3 at once. It would not give breakpoint class, so it tests reciprocal status directly rather than by proxy - which is arguably the better experiment.
| design | retrospective, three-stage, no new sample collection; stage 1 is a go/no-go balance check on an existing database, stage 2 links outcome through referring study groups, stage 3 validates the reciprocal-status label by RNA |
|---|---|
| readout | stage 1: chi-square of breakpoint class against age group and lineage; stage 2: EFS, OS and end-of-induction MRD by arm; stage 3: measured misclassification rate in each arm |
| controls | 48 downstream-of-exon-2 patients as an architecture control; EPS15 and ELL spliced classes as independent-partner direction checks; AFF1 (1 spliced of 1,262) as a null-partner pipeline check; parallel direct-junction assay in every stage-3 sample so a negative reciprocal result is interpretable |
| cost tier | low |
| timeline | stage 1 days; stage 3 weeks; stage 2 months, and stage 2 is the only stage that needs new agreements |
| power note | roughly 80% power for an absolute difference of about 19 percentage points on a binary endpoint at two-sided alpha 0.05 (this graph's calculation). A null is a bounded null and must be reported as one. |
| data holder | Diagnostic Center of Acute Leukemia, Institute of Pharmaceutical Biology, University of Frankfurt (R. Marschalek); outcome data with the referring national study groups |
| public alternative | call reciprocal junction reads from existing KMT2A-r RNA-seq (TARGET, St. Jude Cloud, GEO series attributable by !Series_pubmed_id) in samples with known outcome |
arms — the numbers behind the chart
{
"reciprocal_negative": 153,
"reciprocal_positive": 101,
"architecture_control": 48
}- A retrospective comparison across two decades of protocols can produce a survival difference from treatment era alone. Stage 1 must balance era as well as age and lineage, and the design as written does not yet say how.
- If stage 1 shows imbalance and the study is abandoned, that non-result should still be published or recorded, or the next group will spend the same effort discovering the same thing.
- Does the Frankfurt archive hold, or can it recover, outcome data for the 153 and 101? Everything downstream of stage 1 depends on an answer nobody has published.
- Do standard fusion callers retain reciprocal junction reads, or discard them as artefacts of the direct call? If they discard them, the public computational route needs a custom caller and stops being cheap.
- Should the 48 architecture-control patients be analysed first? They differ from the index arm more profoundly than the 101 do, and a difference there would be easier to detect and would calibrate the whole design.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Read in full text (PMC10169636): about 50% of MLLT1 breakpoints (n = 241 of 460) localise upstream of MLLT1 exon 1 with no reciprocal transcript expressed; the spliced class for other partners is EPS15 12 of 51, ELL 10 of 119 and AFF1 1 of 1,262; MLLT1 cases are 242 infant, 129 paediatric, 84 adult and 5 unclassified; KMT2A breakpoint preference in ALL switches from intron 11 to intron 9 at about six months of age; and the authors state that the database offers the possibility to analyse specific patient cohorts and that they will share these data with requesting scientists.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
Read in full text via Europe PMC fullTextXML. A diagnostic archive is not an outcome registry: nothing in this paper says whether survival data are linked to the breakpoint records, and the referring study groups are not enumerated.
Evidence that argues against this (1)
- The recombinome is a referral series across two decades and many protocols, so any outcome comparison drawn from it carries era and treatment heterogeneity that a randomised design would not. Attarbaschi A et al. 2023
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Read in full text (PMC11246743): of 302 t(11;19) patients, 50.7% (n = 153) break 5' of MLLT1 and 3' of ACER1, 33.4% (n = 101) break within MLLT1 intron 1 and 15.9% (n = 48) break downstream of exon 2; patient age group is not significantly correlated with YEATS-domain retention status (Chi-square p = 0.33, df = 2) while significantly fewer AML than ALL patients retain it (p < 0.0001, df = 1); and the chromosomal breakpoint data were provided by R. Marschalek of the Diagnostic Center of Acute Leukemia, University of Frankfurt.
The ENL YEATS epigenetic reader domain critically links MLL-ENL to leukemic stem cell frequency in t(11;19) Leukemia. · Hu H, Saha N, Yang Y, Ahmad E, et al. · Leukemia 2023What this evidence cannot show
Read via the PMC HTML page after Europe PMC fullTextXML returned HTTP 500 and efetch db=pmc returned HTTP 429. Both chi-square tests are on the 254-versus-48 split, not on the 153-versus-101 split this design depends on, which is precisely why stage 1 exists.
Evidence that argues against this (1)
- The same paper shows that breakpoint position in this partner changes in vivo leukaemogenicity and leukemic stem cell frequency, so breakpoint class is not a neutral label and the arms may differ for reasons other than reciprocal status. Hu H et al. 2023
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
ACER1, the gene immediately 5' of MLLT1 at 19p13.3, appears in the recombinome's supplementary list of the 40 in-frame fused reciprocal translocation partner genes, and MLLT1 is among the direct partners of the complex rearrangements in which those in-frame reciprocal fusions were identified.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
Read from the supplementary Word file in the Europe PMC supplementaryFiles archive for PMC10169636; the table is alphabetical with no patient-level linkage, so the attribution of the ACER1 case to a KMT2A::MLLT1 patient is inference from genomic adjacency, not a published statement. n = 1.
Evidence that argues against this (1)
- One case in 426 complex rearrangements implies a leak rate well under 1% if the attribution holds, which would dilute rather than fabricate an effect. Meyer C et al. 2023
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In a T-ALL with a cryptic t(11;19), MLL sequences distal to the breakpoint were deleted from the genome, precluding formation of a reciprocal ENL/MLL fusion protein despite an intra-MLLT1 breakpoint.
Complex MLL rearrangement in a patient with T-cell acute lymphoblastic leukemia. · Chervinsky DS, Sait SN, Nowak NJ, et al. · Genes Chromosomes Cancer 1995What this evidence cannot show
Abstract only; no PMCID. n = 1, 1995 Southern and FISH methods. The rate of this configuration in KMT2A::MLLT1 has never been counted.
Evidence that argues against this (1)
- The recombinome catalogues only two patients with the converse configuration across 426 complex rearrangements, suggesting deletion-driven misassignment is uncommon, though no MLLT1-specific count exists. Meyer C et al. 2023
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
KMT2A::MLLT1 is independently associated with adverse outcome in childhood KMT2A-rearranged AML, while t(11;19)(q23;p13.3) T-ALL had the highest 5-year EFS in a 629-patient childhood non-infant KMT2A-rearranged ALL cohort at 91.2% (n = 34) against a whole-cohort 69.1%.
Outcomes of Childhood Noninfant Acute Lymphoblastic Leukemia With 11q23/KMT2A Rearrangements in a Modern Therapy Era: A Retrospective International Study. · Attarbaschi A, Möricke A, Harrison CJ, Mann G, et al. · J Clin Oncol 2023What this evidence cannot show
Abstract-level reading; the AML half of this claim rests on a different paper (PMID 38621200) and the juxtaposition is a cross-study one, not a fitted interaction. n = 34 in the favourable cell, not pre-specified.
Evidence that argues against this (1)
- The T-lineage advantage may be a lineage effect that any partner would show; the paper gives no t(11;19) B-ALL comparator to exclude it. Attarbaschi A et al. 2023
the protocol that would run the comparison
stage 2 is the hypothesis's first partner-matched clinical test
same stratification applied to trial data rather than to a diagnostic archive
designed to close the measured absence
names the age band in which the stratified analysis must be run and shows no published figure exists there