YWHAG / 14-3-3 gamma (P61981) - the only cross-laboratory reproduced KMT2A proximity partner in the curated record, and UniProt places it nowhere near the nucleus

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.

• 14-3-3 gamma is one of seven paralogues with overlapping clients, and DepMap gives -0.2275 with only 6.0% of lines below -0.5. Single-paralogue inhibition would be compensated, which is the same paralogue-redundancy failure this graph has documented for the KDM3 and KDM6 families.
• 14-3-3 proteins bind a very large number of phosphoprotein clients across general and specialised signalling pathways. An inhibitor would be a hub inhibitor, and the off-target surface is the phosphoproteome rather than a gene set.
• YWHAG loss-of-function in humans causes developmental and epileptic encephalopathy 56, with refractory seizures and poor prognosis after normal early development. An intervention in an infant population would be acting on a gene whose impairment in infancy is already known to produce that.
• If the association really is bridged by an AFF-family scaffold, as this graph's standing hypothesis holds, then targeting 14-3-3 would not touch the fusion at all - it would remove a chaperone from a complex that has already formed. The bite-back is that the intervention would be aimed at the wrong molecule in the chain.

1. Is there a nuclear pool of 14-3-3 gamma in any cell? UniProt annotates none, and the whole cross-study proximity observation presupposes one. A single immunofluorescence or fractionation experiment in a t(4;11) line would settle it, and it is the cheapest discriminating experiment this entry suggests.
2. Read PMID 40022573, the large BCP-ALL cell-line resource paper that names YWHAG in its abstract. It is B-lineage ALL, it is unread, and if it reports YWHAG expression or dependency across those lines it is the most directly relevant datum available.
3. Are 14-3-3 paralogue peptides distinguishable by mass spectrometry in these datasets? If not, the attribution of the reproduced recovery to gamma rather than to the family is unsafe, and the 'only cross-study reproduction in the curated record' claim weakens to a family-level one.
4. Does AFF1/AFF4 depletion abolish 14-3-3 recovery of KMT2A? This graph's standing hypothesis says the association is bridged by an AFF-family scaffold and specifies the experiment (exp-14-3-3-bridge-pulldown-t411); the absence of a nuclear gamma pool recorded here is a new independent argument for it.
5. Should this graph hold entries for the other five paralogues recovering KMT2A - YWHAB, YWHAE, YWHAH and SFN? Six of seven recovered it in one study; writing one and not the others records the bait-selection accident of the second study as though it were a biological distinction.

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

   IntAct holds exactly two rows placing YWHAG (P61981) against KMT2A (Q03164): PMID 36931259, proximity-dependent biotin identification (MI:1314) in HEK293 Flp-In, and PMID 39251607, proximity-dependent biotin identification in K562 - both at IntAct MI-score 0.41, and in both rows the experimental role of YWHAG is bait and the experimental role of KMT2A is prey.

   IntAct PSICQUIC, MITAB2.7 rows for Q03164 (KMT2A), filtered on P61981 · EMBL-EBI IntAct 2026
   ▸What this evidence cannot show

   Queried live 2026-09-22; experimental role, host organism, detection method and MI-score read out of MITAB2.7 columns 19/20, 29, 7 and 15 rather than assumed. THESE TWO PUBLICATIONS ARE GENUINELY INDEPENDENT - two groups, two questions, two host cells - which is what distinguishes this row pair from the BioPlex release pairs elsewhere in the read queue. MI-score 0.41 is the second-highest in the queue and still below IntAct's 0.6 conventional high-confidence cut. HOSTS: HEK293 Flp-In and K562. K562 IS the only haematopoietic host anywhere in the fifteen-member read queue - but it is a CML-derived erythroleukaemia line, NOT KMT2A-rearranged and NOT B-lineage, so it does not make the row about the index disease. DIRECTION: YWHAG is the bait in both rows; nobody has tagged KMT2A and recovered YWHAG. PROXIMITY LABELLING MEASURES PROXIMITY, NOT CONTACT: a BioID hit at 10 nm radius is compatible with no physical interaction. Rule 93 endpoint caveat: PSICQUIC returns 218 rows for KMT2A against 376 from IntAct REST, so this is a floor.

   ▸Evidence that argues against this (1)

   • 14-3-3 proteins bind a very large number of partners by recognition of phosphoserine/phosphothreonine motifs, and UniProt's own function line for this paralogue says it is 'implicated in the regulation of a large spectrum of both general and specialized signaling pathways' and 'binds to a large number of partners'. A 14-3-3 hit is therefore intrinsically low-specificity, and reproducing across two 14-3-3-centred or abundance-sensitive proximity datasets is the expected behaviour of a promiscuous hub rather than evidence of a particular relationship. UniProt Knowledgebase 2026
2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   A proximity survey assigning 14-3-3 proteins a central chaperone-like role in human cells recovers KMT2A as prey of six family members - YWHAB, YWHAE, YWHAG, YWHAH, YWHAQ and SFN - and independently YWHAG recovers KMT2A in the K562 50-RBP BioID2 map from a different laboratory, which this graph records as the only cross-study, cross-laboratory reproduction of a KMT2A proximity relation in the entire curated record; while the one DEMONSTRATED 14-3-3 contact in this system is on the partner side, 14-3-3 theta binding AFF1 Ser588 and forming a complex with MLL-AF4 in t(4;11)-positive RS4;11 and MV4-11 cells.

   A central chaperone-like role for 14-3-3 proteins in human cells. · Segal D, et al. · Mol Cell 2023
   PMID 36931259DOI
   ▸What this evidence cannot show

   TWO SOURCES SUMMARISED IN ONE ITEM and the split matters. The six-paralogue recovery and the cross-study reproduction are this graph's own prior finding, recorded in ana-kmt2a-14-3-3-proximity-two-studies and re-checked by this beat against the live IntAct rows for YWHAG (both present, both MI-score 0.41). The AFF1 Ser588 half is from PMID 31493143 (Fioretti et al.) via this graph's interface-ywhaq-aff1-s588-14-3-3 entry, and THIS BEAT DID NOT OPEN THAT PAPER - it is carried from an existing entry that cites it as read, and is flagged as such rather than presented as verified here. The load-bearing point for THIS entry is the contrast: the reproduced relation is gamma's and is proximity-only, while the demonstrated functional contact is theta's and is on AFF1.

   ▸Evidence that argues against this (1)

   • Six of seven paralogues recovering KMT2A in one study is the signature of a family-level or abundance-level effect rather than a specific relationship - if six paralogues do it, the relation is not about gamma. The cross-study reproduction picks gamma out only because gamma happened to be one of the 50 baits in the second study; had the second study used theta, theta would be the reproduced one. The reproduction is therefore partly a property of bait selection. Khoroshkin M et al. 2024
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   UniProt resolves gene_exact:YWHAG in reviewed human records to exactly one entry, P61981 (1433G_HUMAN, 247 aa, '14-3-3 protein gamma'), and its COMPLETE set of subcellular-location comments is Cytoplasm/cytosol (ECO:0000269, PubMed 22532927) and Mitochondrion matrix (ECO:0000269, PubMed 22532927) with a note that it translocates to the mitochondrial matrix following induction of mitochondrial protein catabolism - there is NO nucleus comment, coded or uncoded; the record also carries a disease comment for developmental and epileptic encephalopathy 56 (DEE56), an autosomal disorder with refractory seizures and neurodevelopmental impairment after normal early development, and 22 PDB cross-references.

   UniProtKB - P61981 (1433G_HUMAN), 14-3-3 protein gamma · UniProt Knowledgebase 2026
   ▸What this evidence cannot show

   Fetched live 2026-09-22. THE ABSENCE OF A NUCLEUS COMMENT IS THE LOAD-BEARING OBSERVATION and it is an absence in a database, not an experiment: UniProt not annotating a nuclear pool is weak evidence that no nuclear pool exists, particularly for one of the most abundant cytosolic protein families in the cell. Both coded locations come from ONE paper, PubMed 22532927, which this beat did NOT open - per rule 91 the code establishes that a curator read an experiment there, not that it is the founding or best localisation study for this protein, and a mitochondrial-catabolism paper is an odd sole source for a 14-3-3 paralogue's address. The 22 PDB entries were NOT fetched and no structure, residue range or interface is claimed from them. The DEE56 description is UniProt's own text; no clinical paper was fetched. The feature table has NO Domain features, so no 14-3-3 fold boundary is asserted.

   ▸Evidence that argues against this (1)

   • The family demonstrably reaches the nucleus in this exact disease: 14-3-3 theta and AF4 are reported as NUCLEAR interactors in t(4;11)-positive RS4;11 and MV4-11 cells, with theta binding AFF1 Ser588 and complexing with MLL-AF4. So the correct reading is not that 14-3-3 proteins are absent from the nucleus, but that gamma specifically has no curated nuclear address - which is a much narrower claim and is the one this entry makes. Fioretti T et al. 2019
4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   PubMed esearch 2026-09-22: YWHAG[tiab] AND nuclear[tiab] returns 4 records and all four are off-beat on title-check - a cholangiocarcinoma circNOX4/miR-6884-5p/YWHAG axis paper, a 14-3-3 disease-targeting drug-repositioning workflow, a diabetic-kidney tubular-exosome study, and a paper on PML protein regulating cell adhesion; and YWHAG[tiab] AND leukemia[tiab] returns 3 records, also all incidental on title-check - a BCP-ALL cell-line resource paper, the same PML paper, and a post-Chernobyl B-CLL expression-profiling study. No record in either set reports nuclear YWHAG or YWHAG in KMT2A-rearranged leukaemia.

   NCBI PubMed esearch, title-checked bounded queries with querytranslation recorded · NCBI E-utilities 2026
   ▸What this evidence cannot show

   NEITHER OF THESE IS A ZERO - both return records, and every record in both sets was individually title-checked rather than sampled (4 of 4 and 3 of 3). The titles were read; NONE of the seven abstracts was opened, so 'off-beat' is a title-level judgement and one of them could contain a relevant figure - in particular PMID 40022573, a large BCP-ALL cell-line resource paper, is exactly the kind of omics resource in which a YWHAG measurement could sit in a supplementary table without reaching the title. Each operand separately tagged, no `(A OR B)[tiab]` form. NAME LIST: YWHAG only - '14-3-3 gamma' was NOT run as a separate operand and neither was '14-3-3 protein gamma', which is a real gap because the older literature uses the Greek-letter form. This graph's existing ana-14-3-3-kmt2a-direct-binding-bounded-absence ran the alias-expanded family query and found 5 records with one on-beat, so the family-level absence is already bounded there; this item bounds only the gamma-specific nuclear and leukaemia questions. Bounded by titles and abstracts only.

   ▸Evidence that argues against this (1)

   • PMID 40022573, 'Utility of a Large Series of B-Cell Precursor Acute Lymphoblastic Leukemia Cell Lines as a Model System', is a B-lineage ALL resource paper that names YWHAG in its abstract and was not opened. If that resource reports YWHAG expression or dependency across BCP-ALL lines - including any KMT2A-rearranged ones - it would be the single most directly relevant datum for this entry, and it is unread. Tamai M et al. 2025
5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   In DepMap release 26Q2 as served by Open Targets, YWHAG (ENSG00000170027) has a mean gene effect of -0.2275 across 1,258 screens, a minimum of -0.802, 6.0% of lines below -0.5, and a mean of -0.2058 across the 57 leukaemia-annotated screens.

   Open Targets Platform GraphQL, target.depMapEssentiality (DepMap 26Q2) · Open Targets Platform / DepMap 2026
   ▸What this evidence cannot show

   RELEASE RECORDED per rule 83: meta.dataVersion = {year: 26, month: 06} = DepMap 26Q2, fetched 2026-09-22. 1,258 is this gene's own screen count. All 29 tissue blocks concatenated before averaging. Ensembl id resolved through the HGNC primary ensembl_gene_id per rule 78, readback control passed (Open Targets returned approvedSymbol YWHAG). The leukaemia subset is a substring match on diseaseFromSource and is NOT a KMT2A-rearranged subset. THE NUMBER IS UNINTERPRETABLE AS A STATEMENT ABOUT 14-3-3 FUNCTION: there are seven paralogues spanning 245-255 aa with overlapping client repertoires, so a single-gene CRISPR score measures paralogue redundancy as much as it measures requirement. No median-split proportion is reported, per rule 84.

   ▸Evidence that argues against this (1)

   • Paralogue redundancy makes this the least informative kind of DepMap result: a flat score for one of seven functionally overlapping family members is the expected outcome whether or not the family is essential, so it cannot be used either to rule the protein out or to rule it in. The experiment that would be informative is a combinatorial one across paralogues, and DepMap does not contain it. Open Targets Platform / DepMap 2026
6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   Resolved live against reviewed human UniProt records, the seven 14-3-3 paralogues span 245-255 residues - YWHAQ P27348 245, YWHAZ P63104 245, YWHAB P31946 246, YWHAH Q04917 246, YWHAG P61981 247, SFN P31947 248, YWHAE P62258 255 - so YWHAG and YWHAQ are TWO residues apart, far inside the 16-residue limit below which an accession-plus-length check cannot discriminate; and gene_exact:SFN returns TWO reviewed human records, P31947 (14-3-3 sigma, 248 aa) and Q9Y3B8 ('Oligoribonuclease, mitochondrial', 237 aa), a different protein carrying SFN as a gene synonym.

   UniProt REST, gene_exact queries for all seven human 14-3-3 paralogues, reviewed records only · UniProt Knowledgebase 2026
   ▸What this evidence cannot show

   All seven queries run live 2026-09-22 and the protein NAME was read back for every one, which is the only defence available at these length separations. THE CONSEQUENCE IS SPECIFIC TO THIS GRAPH AND IS THE REASON THE ROW MATTERS: this corpus assigns theta (YWHAQ, 245 aa) a demonstrated functional contact with AFF1 Ser588 and complexing with MLL-AF4, and gamma (YWHAG, 247 aa) a proximity-only reproduction against KMT2A. A swap between them would convert a proximity observation into a functional one or move a functional result onto the wrong protein, and verify-accessions.py - which compares accession and length - cannot detect a 2-residue difference. data/symbol-collisions.json (147 rows, 2026-09-22) was grepped for YWHAG, YWHAQ, '14-3-3' and SFN before the first query per rule 88: NO row exists for any 14-3-3 symbol or for the family, so this is a NEW trap. NOT CHECKED: whether the paralogues' SEQUENCES are similar enough that a mass-spectrometry experiment could misassign peptides between them - that is the deeper version of this trap and it was not measured here.

   ▸Evidence that argues against this (1)

   • The trap has a reading that cuts against this whole entry rather than merely against a bookkeeping error. If 14-3-3 paralogues are close enough in sequence that shared peptides are common, then a BioID or AP-MS dataset may not be able to attribute a recovery to gamma rather than to another paralogue at all - in which case 'YWHAG reproduces across two studies' could be 'a 14-3-3 peptide set reproduces across two studies'. This beat did not test paralogue peptide uniqueness and cannot exclude it. UniProt Knowledgebase 2026

• co-occurs with →KMT2A / MLL1 — the gene that breaks

  two INDEPENDENT BioID studies, two labs (36931259 HEK293 Flp-In; 39251607 K562); YWHAG bait, KMT2A prey in both; MI-score 0.41. The only row pair in the read queue that is not one screening programme seen twice

• supports →KMT2A sits in the 14-3-3 proximity neighbourhood, and the YWHAG link reproduces across two independent studies

  the gene entry behind that analysis's cross-study reproduction claim, re-checked against live IntAct rows

• supports →KMT2A is a 14-3-3 client by proxy: the contact is on an AFF-family scaffold, not on KMT2A

  NEW independent argument for the bridging hypothesis: UniProt P61981 annotates NO nuclear pool for 14-3-3 gamma, coded or uncoded, so its recovery of nuclear KMT2A is more likely a property of what it is holding than of its own residency

• related to →14-3-3 theta binds AF4 at Ser588 and rides into the MLL-AF4 complex - the phospho-gated contact the fusion really does carry

  PARALOGUE DISCRIMINATION TRAP: gamma (P61981, 247 aa) and theta (P27348, 245 aa) are TWO residues apart, inside rule 66's floor, and carry non-interchangeable claims here - theta a demonstrated AFF1 S588 contact, gamma a proximity-only KMT2A reproduction

• supports →Nobody has ever shown 14-3-3 binding directly to KMT2A: five records across every paralogue alias, and one of them is about AF4

  gamma-specific extension of that family-level absence: YWHAG AND nuclear = 4 records, all off-beat on title-check; YWHAG AND leukemia = 3, all incidental

• ← supportsIn all 24 read-queue rows inspected, the candidate is the BAIT and this project's protein is the prey - nobody has ever tagged KMT2A, AFF4, menin or PSIP1 and found any of them; plus seven bounded rejections

  both YWHAG rows read per row: candidate bait, KMT2A prey, MI-score 0.41; the K562 row is the only haematopoietic host in the whole queue