NF1 - the only tumour-suppressor loss in this disease, and it is a deletion, so a mutation panel cannot see it

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

• NF1 loss cannot be reversed pharmacologically, so a patient found to carry it gains a prognosis and not a treatment - which is the shape of finding that makes genotyping feel useful while changing nothing.
• Because NF1 loss activates RAS without a RAS codon change, a patient with NF1 deletion is RAS-wild-type on every hotspot panel and would be excluded from a RAS-genotype-selected trial while having the pathway lesion that trial was designed around.
• MEK inhibition in NF1-deficient cells has a well-documented paradox in other tumours - relief of negative feedback on upstream receptors - so downstream inhibition of an upstream brake loss is not the clean substitution it is usually written as.
• Searching the literature for this gene by its symbol retrieves both a family of haematopoietic transcription factors and a clinical syndrome. Any claim built on a bare NF1 literature count in this project should be treated as unverified until the titles were read.

1. How many published KMT2A-rearranged cohorts that report no NF1 lesion ran a copy-number assay capable of detecting one? The answer would convert a frequency claim into a testable one, and nobody has audited it.
2. Is NF1 loss in this disease ever germline? Children with neurofibromatosis type 1 have raised myeloid leukaemia risk; whether any reported somatic NF1 event in KMT2A-rearranged ALL sits on a germline background is not stated in any cohort read for this entry.
3. Would NF1-deleted KMT2A-rearranged cells be more or less menin-inhibitor sensitive? Loss of the RAS brake is a pathway-level activation that no menin-inhibitor resistance study has modelled.
4. Does a homozygous NF1 deletion in this disease arise as part of the complex chromosomal rearrangement that produces the fusion, or independently? Half the structural events in the index cohort were direct consequences of the KMT2A rearrangement.

1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   In whole-genome sequencing of infant MLL-rearranged ALL, NF1 was among the non-recurrent kinase/PI3K/RAS lesions, and the NF1 event in case INF018 was a homozygous deletion - the study's named inactivating, as opposed to activating, pathway lesion; copy-number was assessed by SNP array across the validation cohort.

   The landscape of somatic mutations in infant MLL-rearranged acute lymphoblastic leukemias. · Andersson AK, Ma J, Wang J, et al. · Nature Genetics 2015
   PMID 25730765DOI
   ▸What this evidence cannot show

   One case. The paper's phrasing places the homozygous NF1 deletion in INF018 inside a parenthetical about which of the pathway lesions are activating, so the deletion is stated rather than characterised: no breakpoints, no confirmation assay and no statement of whether the second allele was lost by deletion or by another mechanism are given in the main text. Read from the full text via efetch db=pmc (PMC4553269).

   ▸Evidence that argues against this (1)

   • The public curation of the same study additionally lists an NF1 missense variant V1291I of uncertain significance, so the count of NF1 events in this cohort depends on whether variants of uncertain significance are included - a reminder that a one-case frequency is a curation decision as much as a measurement. cBioPortal curation team 2015
2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   Four human transcription factors of the NFI family - NFIA (Q12857, 509 aa), NFIB (O00712, 420 aa), NFIC (P08651, 508 aa) and NFIX (Q14938, 502 aa) - carry the token NF1 inside their UniProt protein names (NF1-A, NF1-B, NF1-C, NF1-X), while neurofibromin is P21359 at 2,839 aa with no gene synonyms.

   UniProtKB reviewed human entries retrieved by gene_exact:NF1 and by protein_name:"nuclear factor 1", with fields accession, gene_primary, gene_synonym, protein_name and length · UniProt Consortium · UniProt 2026
   ▸What this evidence cannot show

   A database fact about names, fetched 2026-09-20. It establishes that a name collision exists; it does not establish that any particular paper in the leukaemia literature has been mis-assigned because of it, and this entry does not claim one has. The reverse risk also exists: a genuine neurofibromin paper indexed only under 'neurofibromatosis type 1' would be missed by an NF1 symbol query.

   ▸Evidence that argues against this (1)

   • The gene-level query is unambiguous - one reviewed record, no synonyms - so unlike the CSNK2A1/CSNK2A2 and STAG1/PMEPA1 cases already logged in this project, a correctly constructed gene_exact query does not return the wrong protein here. The hazard lives in free-text search and in reading protein names back, not in accession resolution. UniProt Consortium 2026
3. E3weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

   In 488 paediatric B-ALL cases from CCCG-ALL-2015 profiled by whole-transcriptome sequencing, 6 of 11 NF1-mutated patients carried a concurrent FLT3 mutation (phi = 0.16, chi-square 13.0, P < 0.001), FLT3/NF1 co-mutation was independently associated with inferior overall survival (HR 12.39, 95% CI 1.40-109.46, P = 0.024) and with day-46 MRD >= 0.01% (OR 9.50, 95% CI 1.65-54.71, P = 0.01), while RAS-pathway alterations as a class were not significantly associated with inferior survival.

   The clinical relevance of RAS pathway gene mutations in childhood B-cell acute lymphoblastic leukemia. · Chen X, Gao L, Zhang Y, et al. · Annals of Hematology 2026
   PMID 42440143DOI
   ▸What this evidence cannot show

   Eleven NF1-mutated patients in total, so every NF1 statement here rests on single-digit event counts; the OS hazard ratio's confidence interval spans 1.40 to 109.46. Variants were called from whole-transcriptome sequencing, which cannot detect the homozygous deletion that is NF1's documented lesion in KMT2A-rearranged disease, so this cohort's NF1-mutant group is by construction a point-mutation group. The authors state the survival effect was clearest compared with patients WITHOUT KMT2A rearrangements. Single centre, retrospective. Graded weak. Abstract read; full text not retrieved.

   ▸Evidence that argues against this (1)

   • The paper's own headline finding is that RAS-pathway alterations as a class were NOT associated with inferior survival, which contradicts the Interfant result in KMT2A-rearranged infants (HR 3.194) and shows that the prognostic weight of RAS-pathway lesions is not portable across paediatric B-ALL populations. Driessen EM et al. 2013
4. E4weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

   In 43 paediatric ALL and AML patients with high-risk clinical features, RAS-pathway aberrations including N-RAS, NF1 and PTPN11 were frequent somatic mutations and were associated with decreased event-free survival (22.8 vs 5.6 months, P = 0.04) and overall survival (124 vs 22.5 months, P = 0.04).

   Mutations predictive of hyperactive Ras signaling correlate with inferior survival across high-risk pediatric acute leukemia. · Ney GM, Anderson B, Bender J, et al. · Translational Pediatrics 2020
   PMID 32154134DOI
   ▸What this evidence cannot show

   n = 43, single institution, ALL and AML pooled, cohort explicitly selected for high-risk features, and NF1 is one of three genes pooled into the exposure variable, so no NF1-specific effect can be read off it. It is cited here only as evidence that NF1 appears in paediatric acute leukaemia series at all. Abstract read.

   ▸Evidence that argues against this (1)

   • Selection on high-risk features guarantees an adverse comparator group and inflates any association found within it; the 488-patient unselected cohort found no class-level RAS-pathway survival effect. Chen X et al. 2026
5. E_q93_nf1_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   UniProt P21359 carries three subcellular locations - Nucleus and Nucleus/nucleolus on one ECO:0000269 (PubMed:14988005) and Cell membrane on ECO:0000269 PubMed:34626534 - plus a bipartite nuclear localization signal annotated at residues 2555-2571; PubMed:34626534 is a SPRED2 human-genetics paper whose abstract describes no neurofibromin localisation experiment.

   UniProtKB P21359 (NF1_HUMAN), SUBCELLULAR LOCATION comments, NLS feature and TISSUE SPECIFICITY, fetched live 2026-09-21 · UniProt Consortium · Databases 2026
   ▸What this evidence cannot show

   The rule-91 finding here is on the MEMBRANE call, not the nuclear one. SPRED proteins are the known recruiters of neurofibromin to the plasma membrane, so the annotation is mechanistically defensible and may rest on a figure inside the paper; what is recorded is that the paper's abstract is about SPRED2 variants and a recessive Noonan-like phenotype, so a reader checking the pointer finds a different gene's disease study. This claim is bounded by the abstract of 34626534, not its full text.

   ▸Evidence that argues against this (1)

   • The sharper criticism of this record is not the wrong-looking pointer but the FRAGMENT problem, and it applies to the nuclear call UniProt does get from the right paper. Both of P21359's nuclear values come from one study that expressed GFP fusions of exons 28-49 in COS-7 cells. That is roughly a third of a 2,839-residue protein in a monkey kidney line, and a construct carrying an NLS but not the rest of the molecule is close to a positive control for nuclear import. The graph should hold NF1's nuclear pool as real - the independent PML-body imaging in a human line supports it - while noting that the annotation itself rests on a fragment. Vandenbroucke I et al. 2004
6. E_q93_nf1_loc2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   In COS-7 cells, GFP fusions of NF1 exons 28-49 identified a functional nuclear localization signal in EXON 43 whose mutation completely abolishes nuclear entry of the fusion protein, and the authors report that a highly expressed splice variant lacking this NLS controls the localisation and hence the function of neurofibromin.

   Neurofibromin is actively transported to the nucleus. · Vandenbroucke I, Van Oostveldt P, Coene E, De Paepe A, Messiaen L · FEBS Lett 2004
   PMID 14988005DOI
   ▸What this evidence cannot show

   COS-7 monkey kidney cells and a truncated GFP construct, not endogenous full-length neurofibromin and nothing haematopoietic (brief rule 73). The paper is UniProt's source for BOTH the nucleus and the nucleolus values; the nucleolar part of that is not corroborated by the only paper that later imaged the protein subnuclearly. Bounded by the abstract.

   ▸Evidence that argues against this (1)

   • The splice-variant observation cuts against the simple reading of this result. If a highly expressed isoform lacks the NLS, then the nuclear pool is a property of the ISOFORM MIXTURE in a given cell type and not of the gene, and no paper has measured that mixture in a haematopoietic cell, let alone a KMT2A-rearranged one. That makes 'NF1 is partly nuclear' a statement whose truth value is cell-type dependent in a way nothing in this graph can currently resolve - the same shape as MEIS1, where this graph already records that one isoform is cytoplasmic and the nuclear call is a sequence prediction. Godin F et al. 2012
7. E_q93_nf1_loc3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   In the HUMAN CCF ASTROCYTOMA line, immunofluorescence shows a partial nuclear localisation of neurofibromin and partial colocalisation with PML nuclear bodies, and a direct interaction between neurofibromin and the PML multiprotein complex was demonstrated by in-situ proximity ligation assay.

   A fraction of neurofibromin interacts with PML bodies in the nucleus of the CCF astrocytoma cell line. · Godin F, Villette S, Vallée B, Doudeau M, Morisset-Lopez S, Ardourel M, Hevor T, Pichon C, Bénédetti H · Biochem Biophys Res Commun 2012
   PMID 22293200DOI
   ▸What this evidence cannot show

   One laboratory, one cell line, a glial tumour line rather than anything haematopoietic, and never independently replicated - the only follow-up is the same group's own 2024 SUMOylation paper (brief rule 73: the system is human astrocytoma). A PLA signal establishes proximity within roughly 40 nm and is not the same as a binary interaction. 'A fraction' is the authors' own framing, so this is a minority pool, which is the hardest kind of localisation claim to confirm or refute. Bounded by the abstract.

   ▸Evidence that argues against this (1)

   • This result directly CONTRADICTS the compartment UniProt assigns, and the contradiction is not acknowledged anywhere. UniProt says nucleolus, from GFP fragments in COS-7; this paper says PML bodies, from endogenous protein in a human line with PLA support. They are different compartments and both cannot be the principal subnuclear address. Which is more likely to be right is not obvious - the nucleolus is the classic destination of an over-expressed basic fragment, which favours the PML result - but the honest position is that NF1's subnuclear address rests on two single-laboratory experiments that disagree, and nobody has run the experiment that would separate them: endogenous full-length neurofibromin co-stained for PML and for fibrillarin in the same cells. Vandenbroucke I et al. 2004
8. E_q93_nf1_loc4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   Full-length neurofibromin isoform 2 and an isolated SecPH fragment are substrates of the SUMO pathway, with two main modified lysines of which K1731 is highly conserved and surface-exposed; modification of K1731 requires neither the surrounding inverted SUMO consensus motif nor a candidate SIM within SecPH, depending instead on specific structural elements of the SecPH domain, some of which are affected in reported NF1 pathogenic variants.

   An Atypical Mechanism of SUMOylation of Neurofibromin SecPH Domain Provides New Insights into SUMOylation Site Selection. · Bergoug M, Mosrin C, Serrano A, Godin F, Doudeau M, Dundović I, Goffinont S, Normand T, Suskiewicz MJ, Vallée B, Bénédetti H · J Mol Biol 2024
   PMID 39216515DOI
   ▸What this evidence cannot show

   In-vitro SUMOylation plus structural modelling and site-directed mutagenesis; this is biochemistry and structure, not a localisation experiment, and it is cited here for the PML-SUMO link rather than for a compartment. The same laboratory produced the PML-body result it builds on, so the two are not independent. No haematopoietic system. Bounded by the abstract.

   ▸Evidence that argues against this (1)

   • The inferential step this evidence tempts - PML bodies are SUMOylation hotspots, neurofibromin is SUMOylated, therefore the PML localisation is functional - does not follow, and the paper does not make it. SUMOylation of a surface lysine that does not need a consensus motif is consistent with an opportunistic modification anywhere in the nucleus, and the authors' own framing of the significance is about SUMOylation SITE SELECTION rather than about neurofibromin's address. Recorded here as a second, non-independent observation from the one group working on nuclear neurofibromin, not as corroboration of the compartment. Godin F et al. 2012
9. E_q93_nf1_loc5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   Bounded 2026-09-21: the NF1 gene clause against a nine-term KMT2A/MLL fusion clause returns 25 records and the same clause AND a localisation clause AND the fusion clause returns 0; against a twelve-term subnuclear clause the gene clause returns 10 records of which only 1 is about neurofibromin, three being about the unrelated NUCLEAR FACTOR 1 including a paper titled 'Nuclear factor 1 is a component of the nuclear matrix'.

   NCBI E-utilities esearch db=pubmed; querytranslation read back and stored beside every count · NCBI E-utilities · Databases 2026
   ▸What this evidence cannot show

   Titles and abstracts only; supporting figures not searched (brief rule 80). The operand control is 13,313 records, almost all of them clinical neurofibromatosis, so the gene clause here is a very noisy instrument and every count above the load-bearing zero was title-screened by hand rather than trusted.

   ▸Evidence that argues against this (1)

   • The 1-of-10 yield is a warning about this entry's own existing collision note rather than a new finding, and it sharpens it in a specific direction: the collision is not merely that other proteins carry NF1 in their names, it is that ONE OF THEM HAS A PUBLISHED SUBNUCLEAR ADDRESS. A localisation beat that counted without reading would have given neurofibromin the nuclear matrix, sourced to a real paper with a title that reads exactly like the answer. That is the same failure shape as the speckle contamination this project logged for HRP2, AF17 and MOF - a token whose off-topic matches are not noise but confident wrong answers. Godin F et al. 2012

• related to →NF1 in KMT2A-r leukemia — one homozygous deletion, one missense, and the only loss-of-function co-lesion in the disease

  gene-level dossier: accession, the NFI-family name collision, and why a mutation panel cannot see this lesion

• part of →Kinase/PI3K/RAS pathway activation — the only common co-lesion class in KMT2A-r leukemia, and it is mostly subclonal

  the only loss-of-function member of the pathway set the summary entry covers

• co-occurs with →FLT3 / CD135 — the strongest non-tether co-dependency, and a target the field may have written off on a badly executed trial

  6 of 11 NF1-mutated paediatric B-ALL patients carried a concurrent FLT3 mutation (phi 0.16)

• co-occurs with →Infant acute lymphoblastic leukaemia (ALL diagnosed under 1 year of age)

  one homozygous deletion in 22 WGS cases; under-ascertained wherever copy number was not assayed

• ← co-occurs withPIK3R1 in KMT2A-rearranged infant ALL - the only cooperating lesion in this disease that was functionally proved to be activating, and it is an ITD

  case INF018 carried both the homozygous NF1 deletion and the PIK3R1 duplication

• related to →PML nuclear bodies and PML::RARA — the best-worked case of a fusion destroying a nuclear compartment, and what it actually predicts for KMT2A

  A fraction of neurofibromin partially colocalises with PML NUCLEAR BODIES in the human CCF astrocytoma line, with a direct interaction shown by in-situ PLA (PMID 22293200), and the same group later showed full-length isoform 2 is SUMOylated at conserved surface K1731 (PMID 39216515). UniProt instead says nucleolus, from GFP fragments in COS-7 - the two disagree and nobody has run the co-stain that would separate them (Q93 2026-09-21)

• related to →A partner's 'normal' compartment is usually read off a different tissue, and for at least two KMT2A partners that reading is wrong in blood cells

  Worked example in both directions on one gene: UniProt's NF1 membrane call is sourced to a SPRED2 human-genetics paper with no neurofibromin localisation in its abstract, and UniProt's subnuclear call (nucleolus) is contradicted by the only paper that imaged the endogenous protein (PML bodies) (Q93 2026-09-21)

• related to →MEIS1's 'nucleus' is a profile prediction, not a measurement - and a homeodomain-less isoform is cytoplasmic and does the opposite job

  Same isoform trap. NF1's nuclear pool depends on a HIGHLY EXPRESSED SPLICE VARIANT THAT LACKS THE EXON-43 NLS (PMID 14988005), so 'NF1 is partly nuclear' is cell-type-dependent on an isoform ratio nobody has measured in blood - exactly the problem this node records for MEIS1 (Q93 2026-09-21)

• related to →Nuclear matrix and mitotic chromosome scaffold — KMT2A sits with topoisomerase II, in the same place the gene breaks

  A NEGATIVE link, recorded to block an error: NF1[tiab] against a subnuclear clause returns a paper titled 'Nuclear factor 1 is a component of the nuclear matrix' (PMID 8089200), which is about NFI/CTF and NOT neurofibromin. 1 of 10 hits on that query is about the right protein (Q93 2026-09-21)