SGF29 — a SAGA reader with a Tudor pocket, intermediate essentiality, and the best unexploited profile in this beat

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.

• Intermediate essentiality cuts both ways: 58.5% of cell lines are depleted by SGF29 loss, which is close enough to pan-essential that a therapeutic window cannot be assumed.
• Not fusion-selective: the discoverers frame SGF29 as a general non-oncogenic AML dependency, so a SGF29 drug would not be a KMT2A-directed therapy.
• Compensation: SGF29 reads H3K4me3, a mark deposited by six different complexes, so blocking the reader does not remove the mark and other readers remain.
• Missing dossier: no verified mouse knockout and no verified expression data in this entry. Those gaps are toxicity questions that have not been asked, not toxicity questions that were answered reassuringly.

1. Does SGF29's near-miss co-dependency with MEN1 (r = 0.24, CI touching zero) become significant in a larger leukemia panel? SGF29 is the single best candidate in this beat for retesting, because it combines intermediate essentiality with a leukemia-enriched mean and a druggable pocket.
2. Is SGF29 required in KMT2A::AFF1 B-ALL, not just in AML? Every published leukemia result is myeloid, and the two most depleted lines in DepMap's leukemia set are B-ALL lines, which nobody appears to have followed up.
3. Does occupying the Tudor pocket phenocopy SGF29 deletion? The genetic result is deletion; a reader-pocket ligand blocks only the H3K4me3 interaction and leaves the rest of SGF29 in SAGA.
4. What is SGF29's normal-tissue expression profile and its mouse knockout phenotype? Both are unfilled in this entry and both are cheap to obtain.

1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   SGF29 binds histone H3K4me2/3 through a tandem-Tudor domain and is required for SAGA complex recruitment to chromatin and for histone H3 acetylation.

   Sgf29 binds histone H3K4me2/3 and is required for SAGA complex recruitment and histone H3 acetylation. · Bian C, Xu C, Ruan J, Lee KK, Burke TL, Tempel W, Barsyte D, Li J, Wu M, Zhou BO, Fleharty BE, Paulson A, Allali-Hassani A, Zhou JQ, Mer G, Grant PA, Workman JL, Zang J, Min J · EMBO Journal 2011
   PMID 21685874DOI
   ▸What this evidence cannot show

   Structural and biochemical work in yeast and human systems; not a leukemia paper and not about KMT2A fusions.

   ▸Evidence that argues against this (1)

   • Required for SAGA recruitment and H3 acetylation is SAGA biology. The connection to a KMT2A fusion is that both act at H3K4me3-marked promoters, which is a shared address rather than a shared mechanism. Cell Reports 2014
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   A CRISPR domain-focused screen for MEIS1 regulators identified SGF29; SGF29 deletion impairs leukemogenesis across multiple AML subtype models, and the authors propose the SGF29 Tudor domain as a drug target while framing SGF29 as a non-oncogenic dependency.

   Transcriptional control of leukemogenesis by the chromatin reader SGF29. · Barbosa K, Deshpande A, et al., Deshpande AJ · Blood 2024
   PMID 38048593DOI
   ▸What this evidence cannot show

   AML models rather than KMT2A::AFF1 infant B-ALL; the prototype system is AF10-rearranged leukemia. CRISPR deletion is not pharmacological Tudor-pocket blockade, and whether occupying the reader pocket reproduces the deletion phenotype is untested. The authors' own framing as a general AML dependency, rather than a fusion-specific one, is a caution about selectivity.

   ▸Evidence that argues against this (1)

   • The same group's follow-up locates SGF29's essential output in ribosomal gene expression, which makes it a housekeeping dependency rather than an oncogene-specific one. A Tudor-domain-focused CRISPR screen identified SGF29 as crucial for H3K9 acetylation, RIBOSOMAL GENE EXPRESSION and leukemogenesis, and the lead Tudor-domain inhibitor derived from it is active against leukemia. Ribosome biogenesis is required by every proliferating cell including haematopoietic progenitors, so 'non-oncogenic dependency' cuts both ways: it is exactly the profile of a target with no therapeutic index. Note also that this is not independent replication of E2 -- the two studies share authors and the same CRISPR-screen platform -- so the SGF29 case currently rests on one group. Chan AKN et al. 2024
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   SGF29 is not DepMap common-essential yet has intermediate essentiality: mean Chronos -0.567 across 1,258 lines with 58.5% of lines depleted, and a leukemia mean of -0.741 across 57 lines, with NALM-6 (-1.69), AML-193 (-1.41) and OCI-AML2 (-1.31) most depleted; gnomAD reports pLI 0.490 and LOEUF 0.606 with 15 observed against 38.1 expected loss-of-function variants.

   DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL API; gnomAD v4 constraint via the gnomAD GraphQL API · Broad Institute DepMap; gnomAD; Open Targets Platform · Databases 2026
   ▸What this evidence cannot show

   DepMap release version not stated by the API. 58.5% of lines depleted is high enough that SGF29 sits close to the boundary of common-essential classification, and a different threshold would place it on the other side. No HPA expression figure is quoted in this entry because our first HPA query used an identifier that resolved to ERG; we did not re-run it against the correct id, so SGF29 tissue expression is unverified here.

   ▸Evidence that argues against this (1)

   • Intermediate essentiality is being read as a therapeutic window when it may be the opposite. SGF29 is depleted in 58.5% of 1,258 lines, and the mechanism the same group reports for the leukemia requirement is H3K9 acetylation at ribosomal genes. A gene that more than half of all cell lines need, acting on ribosome biogenesis, is a proliferation dependency shared with every haematopoietic progenitor, and the leukemia-versus-rest difference (-0.741 against -0.567) is modest. 'Best unexploited profile in this beat' is a relative statement inside a beat where the alternatives are pan-essential; it is not evidence of a window. Chan AKN et al. 2024
4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   Crystal structures show SGF29 bound to an H3K4me3 peptide at 1.37 A (3ME9) and to H3R2me2s/K4me3 at 1.28 A (3MEU), defining the reader pocket directly.

   RCSB PDB entries 3ME9 and 3MEU, verified against the RCSB data API for title, experimental method and resolution · RCSB Protein Data Bank · RCSB PDB 2026
   ▸What this evidence cannot show

   These are isolated Tudor-domain complexes with short peptides, not SGF29 within SAGA on a nucleosome; pocket geometry in the full assembly may differ.

   ▸Evidence that argues against this (1)

   • Counting PDB entries measures how tractable a protein was to CRYSTALLISE, which correlates with being small, ordered and well-behaved rather than with being druggable. What decides design value is whether the deposited density covers the site that matters, at a resolution that supports modelling, in the relevant state — and this graph has repeatedly found structures described in the direction that overstates what is resolved. Two structures of one tandem-Tudor reader with peptides define the pocket and leave the selectivity question — many proteins read H3K4me3 — untouched. RCSB PDB 2026
5. E5weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.dataset

   Across 54 DepMap leukemia lines SGF29 correlates with MEN1 at r = 0.241 with a 95% confidence interval of -0.005 to 0.458 (permutation p = 0.080), the closest near-miss in the panel, failing Holm correction.

   Leukemia-restricted co-dependency with MEN1, bootstrap and permutation testing; analyses/wrad-discrimination/ in this project · Broad Institute DepMap; Open Targets Platform; analysis run by this project · Database and derived analysis 2026
   ▸What this evidence cannot show

   Our own computation in 54 lines, not significant. A confidence interval that touches zero is a null result and must not be reported as a positive one. It is recorded here only to identify SGF29 as the best candidate for retesting with more lines.

   ▸Evidence that argues against this (1)

   • The near-miss should be read against the deeper null, which is harsher than the 16-gene panel this figure comes from. Re-run against a 2,743-gene variance-matched pool with Holm correction across the same 54 leukemia lines, only KMT2A, PSIP1, DOT1L, FLT3, MEF2C and KAT7 survive; MEIS1 at r = 0.468 already fails at Holm 0.067, and DPY30, MLLT1 and BCL2 fail too. SGF29 at r = 0.241 with a confidence interval that crosses zero is far below any of those. Describing it as 'the closest near-miss in the panel' risks reading it as almost-significant when in the corrected framework it is not close at all. Broad Institute DepMap; Open Targets Platform; analysis run by this project 2026
6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   The correct Ensembl gene identifier for SGF29 is ENSG00000176476, confirmed by resolving both SGF29 and its former symbol CCDC101; the identifier ENSG00000157554, which appears alongside SGF29 in some gene lists, resolves to ERG.

   Ensembl REST cross-reference service, xrefs/symbol/homo_sapiens for SGF29, CCDC101 and ERG, queried 2026-09-18 · Ensembl / EMBL-EBI · Ensembl 2026
   ▸What this evidence cannot show

   Gene symbol to identifier mappings change between Ensembl releases; this was resolved on 2026-09-18 and should be re-checked rather than recalled.

   ▸Evidence that argues against this (1)

   • The correction is the content here, and it is the fourth wrong-identifier incident this project has logged. Recording which identifier was wrong and how it was resolved is more durable than recording only the right one. UniProt Consortium 2026
7. E_q91_sgf29_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   UniProtKB Q96ES7 (SGF29_HUMAN, primary gene SGF29, 293 aa, read back live 2026-09-21) carries ONE subcellular location value, 'Nucleus', on a SINGLE ECO:0000250 whose source is P0C606 - i.e. INFERRED BY SIMILARITY, with no human experimental evidence. P0C606 was fetched in the same session and read back: SGF29_RAT, primary gene Sgf29, synonym Ccdc101, 293 aa, whose own single 'Nucleus' value rests on one ECO:0000269, PubMed:17334388 - a study reporting deregulated expression of rat SGF29, a novel component of TFTC/STAGA histone acetyltransferase complexes, in hepatocellular carcinoma, in the context of c-Myc MbI/MbII-dependent STAGA recruitment.

   UniProtKB Q96ES7 SUBCELLULAR LOCATION comments, evidence codes and feature table, fetched live 2026-09-21 (a single ECO:0000250 nucleus call transferred from the RAT orthologue P0C606) · UniProt Consortium · UniProtKB 2026
   ▸What this evidence cannot show

   The rat and human proteins are both 293 aa and the transfer is very likely correct; what is recorded is the evidence chain, which is human-from-rat-from-one-HCC-paper. The rat paper's abstract describes cloning and complex membership and does not, in reachable text, describe a localisation experiment - so the chain may terminate in the same shape found for SMARCD1 in this tranche.

   ▸Evidence that argues against this (1)

   • The defence is that a tandem-Tudor H3K4me3 reader in SAGA has to be nuclear, so the annotation chain does not matter. That defence is weaker for SGF29 than it looks, because the one system in which SAGA subunit localisation has actually been surveyed found that nuclear residence is NOT automatic for a SAGA subunit: in fission yeast, loss of a single subunit, Sgf73, measurably reduced or altered the nuclear fluorescence localisation of seven others including Sgf29. If subunit-level perturbations move SAGA members out of the nucleus in the one organism where anyone looked, then 'it is a SAGA subunit' is not by itself a compartment. Zhou X et al. 2014
8. E_q91_sgf29_loc2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   The only direct subcellular imaging of any SGF29 orthologue: in fission yeast (Schizosaccharomyces pombe), in-situ integrated fluorescence strains were constructed for ALL 19 SAGA subunits and their localisation sorted into four types; deletion of sgf73+ - the bridge connecting the deubiquitination module to the rest of SAGA - not only significantly reduced the nuclear fluorescence localisation of the deubiquitination subunits Ubp8, Sgf11 and Sus1 but also affected that of the acetylation subunits Gcn5, SGF29 and Ngg1 and of the core structural subunit Spt7.

   [Subcellular fluorescence localization analysis of all SAGA subunits in fission yeast (Schizosaccharomyces pombe)]. · Zhou X, Zhou N, Yu Y, Lv H · Yi Chuan 2014
   PMID 24846945DOI
   ▸What this evidence cannot show

   Schizosaccharomyces pombe, published in Chinese in Yi Chuan, read from the English abstract only. The four localisation 'types' are not enumerated in the abstract, so which type Sgf29 falls into is not readable from here. Yeast SAGA and human SAGA/ATAC differ in subunit composition, and SGF29's human paralogous context includes ATAC, which has no yeast equivalent.

   ▸Evidence that argues against this (1)

   • The finding argues against its own transferability as much as for it. If SAGA subunit localisation is contingent on complex integrity, then the human SGF29 compartment depends on which of the TWO complexes it is in - SAGA or ATAC - and fission yeast has no ATAC, so the one available image cannot speak to half of human SGF29. This is the same structural problem SKIC8 has in this tranche, where a subunit shared between PAF1C and SKI has a compartment that is really two compartments; the difference is that for SKIC8 somebody measured both in human cells and for SGF29 nobody has measured either. Zhu B et al. 2005
9. E_q91_sgf29_loc3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

   Counted absence, 2026-09-22, querytranslation read back on every line. SGF29[tiab] = 44 (operand control) and CCDC101[tiab] = 4. No subnuclear compartment: SGF29[tiab] against nuclear speckle, nucleolus, subnuclear and PML body = 0. No fusion link: SGF29[tiab] AND KMT2A[tiab] = 0, CCDC101[tiab] AND MLL[tiab] = 0, SGF29[tiab] AND MLL[tiab] = 1 (an ER-stress paper whose MLL reference is the SET1/MLL family).

   Counted absence with retired-symbol control: SGF29 versus CCDC101, compartment list and MLL/KMT2A intersections, querytranslation recorded, run 2026-09-22 · NCBI E-utilities esearch (db=pubmed), run by Q91 · PubMed E-utilities 2026
   ▸What this evidence cannot show

   Bounded by titles and abstracts (rule 80). The complex phrases 'SAGA' and 'ATAC' were not run and rule 64 requires them for a subunit absence; both are also ordinary-word-contaminated (ATAC collides with ATAC-seq), which is why they need title-checking rather than counting.

   ▸Evidence that argues against this (1)

   • The retired-symbol control REFUTES the expectation it was run under, and that is worth recording as a result rather than discarding. The dispatch brief flagged CCDC101 as SGF29's retired symbol and predicted the older literature would sit under it, on the pattern of WDR61/SKI8 and OTT3/RBM15B. It does not: CCDC101[tiab] returns 4 records in total, the oldest of which is a genome-wide meta-analysis for severe diabetic retinopathy, and the current symbol returns eleven times more. A retired symbol is a hypothesis to test per gene, not a rule - it under-counted SKIC8 more than twentyfold in this same tranche and does nothing at all here, and a beat that assumed the rule would have spent its queries in the wrong place. Grassi MA et al. 2011
10. E_q91_sgf29_loc4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The one documented cell state for human SGF29: it is required for survival of human cells from ER stress. SGF29 knockdown impairs transcription of the ER stress genes GRP78 and CHOP, and besides histone H3K14 acetylation SGF29 is also required for the MAINTENANCE OF H3K4me3 at those genes - a dual role coordinating acetylation and H3K4 trimethylation. The paper states that human SGF29 is a subunit of the histone acetyltransferase module of BOTH the SAGA and the ATAC co-activator complexes.

    A dual role for SAGA-associated factor 29 (SGF29) in ER stress survival by coordination of both histone H3 acetylation and histone H3 lysine-4 trimethylation. · Schram AW, Baas R, Jansen PW, Riss A et al. · PLoS One 2013
    PMID 23894581DOI
    ▸What this evidence cannot show

    Human cells but not haematopoietic and not fusion-bearing; ER stress is not a state this graph currently models in KMT2A-rearranged disease. No localisation content. This is the single record returned by SGF29[tiab] AND MLL[tiab], and its MLL reference is to the SET1/MLL family of methyltransferases rather than to any fusion.

    ▸Evidence that argues against this (1)

    • A reader could take 'SGF29 maintains H3K4me3' as putting SGF29 upstream of the mark and therefore adjacent to KMT2A's function. The direction is the interesting problem: SGF29 is defined as a READER of H3K4me2/3 through its tandem Tudor domain, so a requirement for MAINTENANCE of the mark makes it simultaneously reader and maintainer, which is a feedback loop rather than a linear pathway. For this graph that blocks the simple prediction - if the fusion redistributes H3K4me3, SGF29 does not merely follow the mark, it may help set where the mark persists, and no experiment distinguishes those in any system. Bian C et al. 2011

• regulates →MEIS1 — rate-limiting for MLL leukemia stem cells, a surviving co-dependency, and embryonic-lethal when removed

  SGF29 was identified in a CRISPR screen for MEIS1 regulators; counter-evidence places its essential output at ribosomal genes rather than the fusion programme

• regulates →MEIS1 — rate-limiting for MLL leukemia stem cells, a surviving co-dependency, and embryonic-lethal when removed

  identified in a CRISPR screen for MEIS1 regulators; deletion impairs leukemogenesis

• related to →Plain-English: chromatin and epigenetics — the sticky notes on the DNA, and which ones matter here

  tandem-Tudor reader of H3K4me2/3, the mark the SET1/MLL family writes

• supports →Open ground: the questions in this graph that nobody in the landscape appears to be working on

  intermediate essentiality, leukemia-enriched, druggable pocket, almost nobody working on it

• ← part ofUSP22 — a KMT2A-rearranged dependency that works through MEIS1 protein stability, and is not an interactor of the fusion

  both are SAGA subunits — SGF29 the reader module, USP22 the DUB module

• 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

  Human SGF29's only nuclear evidence is ECO:0000250 transferred from the RAT orthologue P0C606, whose own support is one hepatocellular-carcinoma paper; the only direct imaging of any Sgf29 is in fission yeast (Q91 2026-09-22)

• related to →KAT2A / GCN5 - the SAGA catalytic subunit that reaches this disease through its complex rather than through its own symbol, and whose only chemical tool degrades its paralogue too

  Fission-yeast imaging of all SAGA subunits shows nuclear localisation of the acetylation module - Gcn5/KAT2A, Sgf29 and Ngg1 - is contingent on Sgf73, so 'it is a SAGA subunit' is not by itself a compartment (PMID 24846945) (Q91 2026-09-22)

• ← related toCTR9 — a PAF1C subunit the field assumes is a fusion partner, with no paper that tested it

  Second consecutive tranche in which a chromatin-complex subunit's human compartment turns out to be a RODENT inference: SGF29 human-from-rat with one hepatocellular-carcinoma paper behind it (Q91), CTR9 human-from-mouse with one 1996 SH2-binding cloning paper behind it. Both have identical or near-identical orthologue lengths, so the length check protects neither (Q93 2026-09-21)

• ← related toKAT2A / GCN5 - the SAGA catalytic subunit that reaches this disease through its complex rather than through its own symbol, and whose only chemical tool degrades its paralogue too

  Same complex, opposite annotation quality, and the contrast is the finding: SGF29's human nuclear call is ECO:0000250 from RAT with one HCC paper behind it (Q91), while KAT2A has FOUR human ECO:0000269 citations plus cell-cycle- and damage-scoped pools. Record quality tracks attention paid to a subunit, not the complex it sits in (Q93 2026-09-21)

• ← related toCSNK2A2 - CK2's other catalytic subunit, expressed in every KMT2A-rearranged line and, if anything, slightly favourable to lose

  THIRD cross-species two-step compartment chain found by this beat in two tranches: SGF29 human<-rat<-one HCC paper, CTR9 human<-mouse<-one 1996 cloning paper, CSNK2A2 human<-mouse<-one adipogenesis paper. All three orthologue pairs are identical or near-identical in length, so the length check protects none of them (Q93 2026-09-21)

• ← related toNASP - the soluble histone reservoir, and the only gene in this beat with a real selective difference at a pan-essentiality shallow enough to interpret

  THE INVERTING CASE among this beat's three rodent-propagation chains. SGF29 and CTR9 got worse when opened; NASP got BETTER - the mouse record's third citation performed its import experiment in human HeLa cells with an NLS-deletion control, so a strict ECO:0000250 filter would have discarded a real human measurement (Q93 2026-09-21)

• ← related toATXN7L3 — the shared adaptor that actually sets H2Bub levels, and a KMT2A-selectivity hit nobody in this field has looked at

  Same complex, opposite modules, and comparable evidence weakness: SGF29's human nuclear call is ECO:0000250 from RAT (tranche 4), ATXN7L3's carries a HAMAP FAMILY RULE plus one ECO:0000269 that is a paper about the paralogue ATXN7L3B. In ATXN7L3's case the apparently misdirected citation is the better one - the paper's point is that ATXN7L3B is cytoplasmic and ATXN7L3 nuclear, and they compete for ENY2 (Q95 2026-09-22)