Planned: read CXCR4 and SYTL1 out of the menin-inhibitor transcriptomes that are already public, in the same contrast that shows HOXA9 and MEIS1 falling
also called CXCR4 menin inhibitor GEO re-analysis
The measurement needed to test whether menin-blocking drugs switch off the marrow hold-still receptor has already been made and uploaded by other laboratories. This is a plan to download it and look.
The same thing, for a scientist
A no-wet-work analysis that scores falsifier F3 of `hyp-cxcr4-function-is-a-hoxa-meis1-output-of-the-fusion`. Take deposited human transcriptomes with a menin-inhibitor arm in KMT2A-rearranged cells - GSE281506 is the confirmed candidate, 42 samples, human, expression profiling by high-throughput sequencing [E1] - verify the perturbation worked by requiring HOXA9 and MEIS1 log2 fold change below -1 at adjusted p < 0.05, then read out CXCR4 and SYTL1 in the identical contrast [E2, E3]. The analysis is worth doing because the corresponding literature is empty: zero papers name menin and CXCR4 together in a title or abstract [E4]. It is status PLANNED and nothing in this entry reports a result.
Design, and the two ways it can fail uninformatively
Design. (1) Resolve every GEO series matching revumenib, ziftomenib, VTP-50469, VTP50469, MI-503, SNDX-5613 or 'menin inhib*' whose organism is human and whose gdstype is expression profiling; retain those with a menin-inhibitor-alone arm in a KMT2A-rearranged background, confirmed from the SOFT family file rather than from the summary page. (2) For each, compute the drug-versus-vehicle contrast. (3) Positive control, and it is the point of the design: require HOXA9 and MEIS1 to fall with log2FC < -1 at adjusted p < 0.05. A series that fails this control is discarded, not reported - the perturbation is the instrument here, and an unvalidated instrument returns a zero that means nothing. (4) Read CXCR4 and SYTL1 in the same contrast. (5) Persist the resolved series ids, sample ids and per-gene statistics, not just the verdict.
Failure mode one: the wrong assay. If MEIS1 controls CXCR4 only through SYTL1-dependent trafficking, CXCR4 transcript can be flat while receptor availability falls. That is why SYTL1 is scored alongside and why a flat CXCR4 alone must not be reported as a negative.
Failure mode two: confounding by differentiation. Menin inhibition induces differentiation, and CXCR4 changes with differentiation state. A fall in CXCR4 is therefore consistent with the hypothesis and with a trivial alternative, which is why the hypothesis's F5 specificity control is a wet experiment and cannot be substituted by more re-analysis.
| script | analyses/cxcr4-under-menin-inhibition/run.py (NOT YET WRITTEN) |
|---|---|
| status | planned |
| instrument control | A series is only analysed if HOXA9 and MEIS1 both fall with log2FC < -1 at adjusted p < 0.05 in the menin-inhibitor arm. A series failing that is discarded and NAMED in the output as discarded, with its ids, so the denominator is visible. |
| persistence requirement | Persist the resolved GEO series ids, the sample ids in each arm, and the per-gene statistics - not the verdict alone. This project has a recorded case where a DepMap scan stored n_kmt2a_rearranged: 11 and not which lines, making every later line-level question unanswerable without re-running. |
| completeness guard | Record complete_run, name any series that failed to download or parse, and exit non-zero if any did. A partial fetch must suppress the verdict, not shrink it. |
| scores falsifier | F3 of hyp-cxcr4-function-is-a-hoxa-meis1-output-of-the-fusion |
| result summary | NOT RUN. No result is reported in this entry and none should be quoted from it. |
inputs
[ "GEO db=gds, human, expression profiling, terms: revumenib, ziftomenib, VTP-50469, VTP50469, MI-503, SNDX-5613, 'menin inhib*'", "SOFT family files for series composition (not acc.cgi, not esummary alone)", "GSE281506 as the confirmed starting candidate" ]
readouts
[ "HOXA9 log2FC and adjusted p (instrument control)", "MEIS1 log2FC and adjusted p (instrument control)", "CXCR4 log2FC and adjusted p (primary)", "SYTL1 log2FC and adjusted p (primary, trafficking route)" ]
- A positive result here would be correlative and confounded by differentiation, and should not on its own be used to argue against adding a CXCR4 antagonist to a menin inhibitor in a trial design.
- A flat CXCR4 would not be a clean negative, because the documented MEIS1 route to this axis is post-transcriptional.
- Does any deposited series carry a menin-inhibitor-alone arm in a B-LINEAGE KMT2A::AFF1 background, or is the public data entirely myeloid? If it is entirely myeloid the analysis answers a neighbouring question, not the index one, and must say so.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
GSE281506 exists and is a human expression-profiling-by-high-throughput-sequencing series of 42 samples titled 'Combining Menin and MEK inhibition to target poor prognosis KMT2A-rearranged RAS pathway-mutant acute myeloid leukemia'; a GEO db=gds query returns 88 records for 'revumenib' and 259 for 'menin inhibitor'.
GSE281506 - Combining Menin and MEK inhibition to target poor prognosis KMT2A-rearranged RAS pathway-mutant acute myeloid leukemia (expression profiling by high throughput sequencing, Homo sapiens, 42 samples) · NCBI Gene Expression Omnibus 2026What this evidence cannot show
Accession, gdstype, sample count, taxon and title read back by esummary on 2026-09-22; samples not inspected. The free-text GEO counts are loose upper bounds on relevance, not counts of usable series. Whether this particular series carries a menin-inhibitor-alone arm is unverified and is step 1 of the analysis.
Evidence that argues against this (1)
- GEO metadata has produced three independent false negatives for this project on a single SuperSeries, one of them a reCAPTCHA page that parsed cleanly as data. The SOFT family file is the authoritative route and the design specifies it for exactly this reason. NCBI Gene Expression Omnibus 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Menin-inhibitor monotherapy down-regulates HOXA9/MEIS1 in KMT2A-rearranged AML, which makes those two genes a usable in-data positive control for whether the perturbation worked in any given deposited series.
Repressing HIF-1alpha-induced HDAC9 contributes to the synergistic effect of venetoclax and MENIN inhibitor in KMT2Ar AML · Ling Q, Zhou Y, Qian Y, et al. · Biomark Res 2023What this evidence cannot show
Myeloid KMT2A-rearranged AML (rule 73). The abstract does not give the magnitude of the HOXA9/MEIS1 fall, so the log2FC < -1 threshold in the design is a reasonable convention rather than a value taken from this paper, and a series failing it might have a real but smaller effect. The threshold should be reported alongside any result.
Evidence that argues against this (1)
- Using the drug's canonical targets as the control means a series in which menin inhibition worked through some other route would be discarded as a failed instrument. That is the conservative choice here, but it is a choice and it biases the analysis toward the textbook mechanism. Ling Q et al. 2023
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
SYTL1 must be read out alongside CXCR4 because MEIS1's documented route to CXCL12/CXCR4 signalling is transactivation of SYTL1, which controls CXCR4 intracellular trafficking rather than CXCR4 transcription.
MEIS1-mediated transactivation of synaptotagmin-like 1 promotes CXCL12/CXCR4 signaling and leukemogenesis · Yokoyama T, Nakatake M, Kuwata T, et al. · J Clin Invest 2016What this evidence cannot show
Murine myeloid HOXA9/MEIS1 models, not a KMT2A fusion and not B-lineage (rule 73). Whether SYTL1 is a MEIS1 target in a human B-precursor KMT2A::AFF1 cell is itself unmeasured, so a flat SYTL1 in human data would be ambiguous between 'not a target here' and 'not moved by the drug'.
Evidence that argues against this (1)
- If SYTL1 turns out not to be MEIS1-dependent in human B-lineage cells, the analysis loses its second readout and reduces to a CXCR4 transcript measurement, which E3's own mechanism says may be the uninformative one. Yokoyama T et al. 2016
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
The corresponding literature is empty: menin[tiab] AND CXCR4[tiab] returns 0 records with valid translation and an explicit 'No items found.', against operands of 1,264 and 17,210, with an identical-shape positive control (menin[tiab] AND HOXA9[tiab]) returning 33.
NCBI E-utilities esearch, db=pubmed - bounded absence with querytranslation readback and rule-67 positive control · NCBI PubMed 2026What this evidence cannot show
Bounded over titles and abstracts (rule 80); a CXCR4 row in a supplementary table is invisible to it, which is the reason this analysis exists rather than a reason to doubt the zero. Queried 2026-09-22.
Evidence that argues against this (1)
- An abstract-level zero cannot bound what is inside papers, and this project has a worked case where a clean bounded zero coexisted with the measurement sitting in a supporting figure of a paper already in the corpus. Van Vlierberghe P et al. 2008
scores falsifier F3 on deposited public data
the named closing route for that absence