SMYD3 - a lysine methyltransferase with three claimed histone marks and one robust non-histone substrate, sitting on an HOXA9 feed-forward loop in MLL-AF9 AML and a fatty-acid-oxidation axis in CML stem cells, with zero interventional trials

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.

• THE CELL-LINE EVIDENCE FOR SMYD3 AS A TARGET IS NOT WEAK, IT IS ABSENT. SMYD3's DepMap pan-essentiality is +0.056 across 1,208 lines - positive - and both arms of the KMT2A comparison sit at zero. The paralogue SMYD2 is +0.036. Neither SMYD protein is required by any cell line in the screen, so any SMYD3 rationale has to explain why 1,208 negatives do not count, and the only available explanation (a self-renewal phenotype under metabolic constraint that a rich-medium proliferation screen cannot see) is an argument rather than a result.
• TEN YEARS OF PROBES AND ZERO INTERVENTIONAL TRIALS. ClinicalTrials.gov contains exactly one study naming SMYD3 and it is observational, against 43 for the menin inhibitors through the identical search. An orally bioavailable first-in-class tool compound published in 2016 that has not entered a trial by 2026 is information about the phenotype it produced.
• THE THERAPEUTIC DIRECTION OF THE DOWNSTREAM AXIS IS NOT DETERMINED. SMYD3 loss depletes CML leukaemia stem cells by shutting off CPT1A-mediated fatty-acid oxidation, while this graph's gene-cpt1a entry records that CPT1A INDUCTION sustains a type-I interferon response and augments PD-1 blockade in refractory tumours. Two published directions, opposite signs, and neither context is a KMT2A-rearranged leukaemia. 'Target the SMYD3-CPT1A axis' currently has no sign.
• THE MECHANISM MAY NOT BE THE MARK. SMYD3's best-supported substrate is the cytoplasmic kinase MAP3K2, the enzyme is described in the primary literature as 'largely cytoplasmic', and its histone mark is disputed three ways. A SMYD3 inhibitor would block methylation of MAP3K2 as well, which potentiates Ras/Raf/MEK/ERK - so in a RAS-co-mutated KMT2A-rearranged leukaemia, which this graph documents as a real cooperating-lesion class, the predicted effect on RAS signalling is real and its sign in that context is unknown.
• THE PARALOGUE TRAP COULD INVERT A RECOMMENDATION. SMYD2 (433 aa) and SMYD3 (428 aa) are five residues apart with near-identical protein names and the same EC number, and verify-accessions.py cannot catch a swap. The mouse genetics in MLL-AF9 leukaemia belong to SMYD2; the HOXA9 loop and the LSC metabolic axis belong to SMYD3. An entry or a grant that swaps them would attribute a germline knockout result to the wrong gene.
• THE DIFFERENTIATION EFFECT CUTS BOTH WAYS IN A PAEDIATRIC PATIENT. The SMYD3-HOXA9 loop blocks differentiation, so inhibiting it is a differentiating therapy - and differentiation syndrome is already the dose-limiting toxicity of the menin inhibitors in this disease. A SMYD3 inhibitor combined with a menin inhibitor, which is the obvious trial, would combine two differentiating agents.
• AND THE COMBINATION THAT IS ACTUALLY SUPPORTED IS WITH A DRUG THAT FAILED. The AML result is a combination of SMYD3-loop disruption with DOT1L inhibition, and this graph's drug-pinometostat entry records what happened to DOT1L inhibition clinically. A combination rationale inherits its partner's clinical history.

1. Is SMYD3 a dependency in KMT2A::AFF1 B-lineage leukaemia? The measurement was nearly made: the HOXA9-loop paper's own panel includes MV4-11, the index-fusion line, and reports only that SMYD3 and HOXA9 induction is LOWER there - so the expression was measured and the dependency was not reported. SMYD3[tiab] AND lymphoblastic[tiab] = 0. An shRNA or degrader in two KMT2A::AFF1 B-lineage lines with a differentiation readout is a two-month experiment on a question the field has skipped twice.
2. Which mark does SMYD3 write in a leukaemic cell, genome-wide? Three marks are claimed from three groups, the only KMT2A-family test is a single-locus H3K4me3 ChIP in one line, and no genome-wide SMYD3 occupancy or calibrated mark quantification exists in any KMT2A-rearranged cell. A calibrated ICeChIP plus SMYD3 ChIP-seq in MOLM-13 and MV4-11 would settle both the mark and the B-lineage question at once.
3. Does the -2 phenylalanine rule out the histone sites? The structures say a phenylalanine two residues before the acceptor lysine is what makes MAP3K2 a robust substrate, and neither H3K4 nor H4K5 has one. A peptide array comparing H3K4, H4K5, H4K20 and MAP3K2 K260 under identical conditions is an afternoon and would explain twenty years of disagreement - OR show the preference does not matter at cellular timescales.
4. Is the SMYD3 to CPT1A axis a target in KMT2A-rearranged disease, and in which direction? SMYD3 loss depletes CML LSCs by shutting down FABP5/PPARD/CPT1A fatty-acid oxidation, and this graph already records that CPT1A INDUCTION potentiates antitumour immunity in a different system - so the sign is context-dependent and neither context is KMT2A-rearranged leukaemia. The quiescent-LSC angle is the one a rich-medium DepMap screen structurally cannot see.
5. Why has no SMYD3 inhibitor reached a trial in ten years? Two chemical series, a covalent probe, an orally bioavailable tool compound, 34 PDB deposits, 342 papers - and one observational study. Either the probes failed to produce a phenotype worth testing, or the phenotype needs a context (LSC self-renewal under metabolic constraint) that standard preclinical models do not supply. Which of those it is determines whether this target is dead or mis-tested.
6. Does SMYD3's centriolar distal-appendage role matter anywhere in this project? SMYD3 is a distal-appendage protein that binds the C2cd3 promoter, and C2CD3 is in this graph as a KMT2A sarcoma partner and a centriole-length gene. Recorded as a convergence with its null stated - C2cd3 is one of five cilia promoters that paper reports - and explicitly not as a pattern.

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

   SMYD3 (Q9H7B4, SMYD3_HUMAN) is a 428-residue histone-lysine N-methyltransferase with a split SET domain at 4-240 interrupted by a MYND-type zinc finger at 49-87 and a C-terminal domain at 272-428 annotated as essential for histone methyltransferase activity, nuclear localization and interaction with HSP90AA1; it has 34 PDB entries, its gene synonyms are ZMYND1 and ZNFN3A1, it carries no DISEASE comment, and its subcellular-location note records that it is mainly cytoplasmic in G0/G1-arrested cells and accumulates in the nucleus at S phase and G2/M.

   UniProtKB Q9H7B4 (SMYD3_HUMAN) - reviewed human record, length 428 aa · UniProt Consortium · UniProtKB 2026
   ▸What this evidence cannot show

   Accession resolved 2026-09-22 by gene_exact:SMYD3 AND organism_id:9606 AND reviewed:true with gene_primary and length read back (rules 62, 65, 66): exactly one reviewed human row, primary gene name SMYD3, entry SMYD3_HUMAN, 428 aa, synonyms ZMYND1 and ZNFN3A1. THE PARALOGUE HAZARD IS THE REASON THE LENGTH IS ASSERTED AND IS NOT SUFFICIENT: SMYD2 is Q9NRG4 at 433 aa, FIVE residues from SMYD3's 428, both named 'SET and MYND domain-containing protein', both carrying EC 2.1.1.354 - so verify-accessions.py, which asserts accession plus length, CANNOT catch a SMYD2-for-SMYD3 swap. That is tighter than the SMC1A/SMC3 sixteen-residue case this project calls the discriminator's limit. UniProt's FUNCTION comment names H3K4 di/tri-methylation and H4K5 methylation (ECO:0000269:15235609, 22419068) and does NOT mention H4K20me3, MAP3K2 or VEGFR1 - so this record carries two of the three claimed histone marks and none of the non-histone substrates.

   ▸Evidence that argues against this (1)

   • The record's own FUNCTION comment is incomplete in a way that matters: it omits the MAP3K2 K260 substrate, which is the one the structural literature calls robust and the one with a mouse genetic requirement, and it omits the H4K20me3 preference reported from the full-length crystal structure. A beat reading only UniProt would conclude SMYD3's substrates are settled and histone-only. Mazur PK et al. 2014
2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   SMYD3 is over-expressed in the majority of colorectal and hepatocellular carcinomas, its introduction into NIH3T3 cells enhances growth while siRNA knockdown suppresses growth in cancer cells, it forms a complex with RNA polymerase II through an interaction with the RNA helicase HELZ and transactivates oncogenes, homeobox genes and cell-cycle genes, it binds the motif 5'-CCCTCC-3' in target promoters such as Nkx2.8, and its SET domain shows histone H3 lysine 4-specific methyltransferase activity that is enhanced by HSP90A.

   SMYD3 encodes a histone methyltransferase involved in the proliferation of cancer cells. · Hamamoto R, Furukawa Y, et al. · Nat Cell Biol 2004
   PMID 15235609DOI
   ▸What this evidence cannot show

   The FOUNDING paper and the ECO:0000269 source for most of UniProt's SMYD3 record, including the H3K4 specificity, the DNA motif, the HSP90 stimulation, the Pol II/HELZ complex, the tissue specificity and the cell-cycle-dependent localisation. Abstract read; full text not read. Colorectal and hepatocellular cancer cell lines plus NIH3T3 - not leukaemia and not a fusion system. The Pol II interaction is described by UniProt itself as possibly INDIRECT and mediated by HELZ, which is worth carrying: 'SMYD3 is in an RNA polymerase II complex' is a weaker statement than it sounds.

   ▸Evidence that argues against this (1)

   • The H3K4 assignment from this paper has been contested twice from the structural side. A 2011 full-length crystal structure with mutational and biochemical analysis reports 'a previously undetected preference for trimethylation of H4K20', and 2016 co-crystals with non-histone peptides identify the determinant of robust catalysis as a phenylalanine at the -2 position, which neither H3K4 nor H4K5 has. Foreman KW et al. 2011
3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   SMYD3 catalyses histone H4 methylation at lysine 5 (H4K5me), a mark detected in diverse cell types whose formation is attenuated by depletion of SMYD3 protein, and SMYD3-driven cancer cell phenotypes require its enzymatic activity.

   Smyd3 regulates cancer cell phenotypes and catalyzes histone H4 lysine 5 methylation. · Van Aller GS, Reynoird N, et al. · Epigenetics 2012
   PMID 22419068DOI
   ▸What this evidence cannot show

   A short report; abstract read, full text not read. It is the ECO:0000269 source for UniProt's H4K5 clause and is one of THREE claimed histone marks for this enzyme. The depletion-attenuation design shows SMYD3 is required for the mark's steady-state level, not that SMYD3 is the writer at any named locus. SMYD3[tiab] AND H4K5[tiab] returns only 4 records, and that count is a FLOOR rather than a measurement because compact mark spellings are a documented query trap in this project's collision registry (CREBBP AND H3K18ac = 0 against CBP AND H3K18 AND acetyl* = 32).

   ▸Evidence that argues against this (1)

   • H4K5 is a canonical ACETYLATION site, and an H4K5 methylation mark reported once and cited in four records is not an established mark. It also competes with the two other claimed marks rather than complementing them - three groups, three marks, on one 428-residue enzyme with a pocket the structural literature calls highly confined. Foreman KW et al. 2011
4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   The crystal structure of full-length human SMYD3 with a SAM analogue shows a split-SET domain in a canonical SET fold closely assembled with a Zn-binding MYND domain and a C-terminal superhelical nine-alpha-helical bundle, whose interlocked arrangement imposes a highly confined binding pocket for histone substrates, and mutational and biochemical analysis established a previously undetected preference for trimethylation of H4K20.

   Structural and functional profiling of the human histone methyltransferase SMYD3. · Foreman KW, Brown M, et al. · PLoS One 2011
   PMID 21779408DOI
   ▸What this evidence cannot show

   Crystallography plus in vitro biochemistry; abstract read, full text not read. 'Previously undetected preference for trimethylation of H4K20' is the paper's own wording and is the THIRD claimed histone mark for this enzyme. In vitro peptide and histone methylation assays are exactly the assay class in which SET-domain substrate preferences have repeatedly failed to survive into cells, so this is recorded as a structural and biochemical preference and not as a cellular mark. SMYD3[tiab] AND H4K20[tiab] = 6, a floor for the reasons at E3.

   ▸Evidence that argues against this (1)

   • The H4K20me3 preference has not displaced the H3K4 assignment in practice: SMYD3[tiab] AND H3K4[tiab] returns 46 records against 6 for H4K20, and the functional test in the system this project cares about - MLL-AF9 AML - used H3K4 methylation as the readout and the catalytically dead F183A mutant failed to rescue. In an MLL-AF9 cell the H3K4me3 reading is the one that was tested and it held. Zhang L et al. 2025
5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   SMYD3 - described in this paper as the largely CYTOPLASMIC lysine methyltransferase SMYD3 - methylates MAP3K2 at lysine 260, which blocks binding of the PP2A phosphatase complex and potentiates Ras/Raf/MEK/ERK signalling; abrogating SMYD3 catalytic activity inhibits tumour development in mouse models of pancreatic ductal adenocarcinoma and lung adenocarcinoma driven by oncogenic Ras, and SMYD3 depletion synergises with a MEK inhibitor to block Ras-driven tumorigenesis.

   SMYD3 links lysine methylation of MAP3K2 to Ras-driven cancer. · Mazur PK, Reynoird N, et al. · Nature 2014
   PMID 24847881DOI
   ▸What this evidence cannot show

   Mouse genetic models of Ras-driven pancreatic and lung adenocarcinoma plus cancer cell lines and protein arrays - NOT leukaemia and NOT a fusion system, stated in the claim per this project's rule 73. Abstract read; full text not read. The phrase 'largely cytoplasmic' is the paper's own characterisation of SMYD3 and is a material qualification on reading SMYD3 as a chromatin writer. This is the substrate the structural literature calls robust and it is the only SMYD3 substrate with a mouse genetic requirement behind it. NOTE: UniProt's FUNCTION comment does not mention MAP3K2 at all.

   ▸Evidence that argues against this (1)

   • A cytoplasmic kinase substrate is very weak support for a nuclear histone activity and, read strictly, competes with it: if the enzyme is largely cytoplasmic and its best substrate is a MAP kinase kinase, the chromatin reading needs the nuclear pool to be doing something the cytoplasmic pool is not. The HSP90-dependent chromatin association and the S/G2/M nuclear accumulation are what supply that, and both are single-paper results. Brown MA et al. 2015
6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   Co-crystal structures of SMYD3 with VEGFR1 and MAP3K2 peptides show that MAP3K2 is a robust substrate because of a phenylalanine at the -2 position relative to the acceptor lysine, accommodated by a shallow hydrophobic pocket on SMYD3 that promotes efficient catalysis, while SMYD3 displays weak activity toward a VEGFR1 peptide because the acceptor lysine sits in the folded kinase domain and would require drastic conformational rearrangement.

   Structural Basis for Substrate Preference of SMYD3, a SET Domain-containing Protein Lysine Methyltransferase. · Fu W, Liu N, et al. · J Biol Chem 2016
   PMID 26929412DOI
   ▸What this evidence cannot show

   Crystallography plus enzymology on peptides; abstract read, full text not read. The paper describes SMYD3's REPORTED substrates as including 'histones (H3K4 and H4K5)', VEGFR1 K831 and MAP3K2 K260 - i.e. it treats the histone assignments as reported rather than testing them, and its own structures are with non-histone peptides. ONE INFERENCE IS THIS BEAT'S OWN AND IS LABELLED AS SUCH, NOT AS THE PAPER'S: neither H3K4 nor H4K5 has a phenylalanine at the -2 position (histone H3 begins ARTK, so K4's -2 is R2; histone H4 begins SGRGK, so K5's -2 is G2), so the stated determinant of robust catalysis is absent from both claimed histone sites. That is a sequence observation, it is testable on a peptide array, and it is not a claim this paper makes.

   ▸Evidence that argues against this (1)

   • A -2 preference is a preference and not a requirement, and SET-domain enzymes methylate substrates lacking their optimal flanking residues at lower efficiency all the time - which for a chromatin writer acting over hours on an abundant substrate may be entirely sufficient. The inference above weakens the histone case; it does not close it, and the MLL-AF9 functional data cut the other way. Zhang L et al. 2025
7. E7moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   IN HUMAN MLL-AF9-DRIVEN MYELOID LEUKAEMIA, SMYD3 is over-expressed in AML and predicts poor prognosis, SMYD3 mediates H3K4me3 enrichment and independently activates HOXA9 transcription, HOXA9 in turn binds the SMYD3 promoter and enhances its transcription, the resulting SMYD3-H3K4me3-HOXA9 loop exacerbates proliferation and blocks differentiation in AML cell lines and in primary patient cells, and combined disruption of the loop with DOT1L inhibition enhances anti-leukaemia activity in vitro and in vivo.

   The transcription factor HOXA9 induces expression of the chromatin modifier SMYD3 to drive leukemogenesis. · Zhang L, Zhong J, et al. · J Biol Chem 2025
   PMID 40451434DOI
   ▸What this evidence cannot show

   ALREADY IN THIS GRAPH AND READ MORE CLOSELY THERE: gene-hoxa9 evidence E8 carries a full-text reading of this paper (via PMC12226129) with the line-level detail - MOLM-13 and THP-1, the F183A catalytically dead rescue failure, the single-locus HOXA9-promoter ChIP in one line, and the DOT1L-independence arm in U937 (CALM-AF10) and 293T. THIS ENTRY DELIBERATELY DOES NOT RESTATE THAT READING and cites the paper for SMYD3's own side of the axis. The system is human but MYELOID and MLL-AF9-driven, not infant B-lineage KMT2A::AFF1. Abstract read here; the full text was read by the beat that wrote gene-hoxa9.

   ▸Evidence that argues against this (2)

   • COUNTER-EVIDENCE FROM THE SAME PAPER, already recorded in gene-hoxa9: SMYD3 and HOXA9 are increased to a GREATER degree in the MLL-AF9 lines MOLM-13, THP-1 and NOMO-1 than in MV4-11, which carries the index fusion KMT2A::AFF1, and the authors state that regulation of SMYD3 by MLL-AF9 and HOXA9 may have some relative specificity within the MLL-rearranged family. So the one strong KMT2A-family result for SMYD3 is weakest in the fusion this project is about. Bagislar S et al. 2016
   • And the family genetics point at the other paralogue: germline Smyd2 ablation retarded MLL-AF9 leukaemia development in mice with a competitive disadvantage in vitro and in vivo, while the same study found Smyd2 AND Smyd3 germline deletions dispensable for normal development. No equivalent mouse genetic test of Smyd3 in MLL-AF9 leukaemogenesis has been reported. Bagislar S et al. 2016
8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   The SMYD3-KMT2A link is unreachable from the approved gene symbol: measured 2026-09-22 with one [tiab] tag per term and querytranslation read back, SMYD3[tiab] AND KMT2A[tiab] returns 2 records and BOTH are off topic - a survey of Kdm/Kmt gene families in the mangrove rivulus fish Kryptolebias marmoratus, and a study of deregulated histone methylases and demethylases in prostate carcinoma - while SMYD3[tiab] AND MLL[tiab] returns 5 and SMYD3[tiab] AND HOXA9[tiab] returns 1, which is the paper that establishes the axis.

   NCBI E-utilities esearch over PubMed and GEO DataSets, run 2026-09-22 by Q98 through a wrapper that refuses any response with an empty querytranslation · Oncofusion knowledge-graph project (Q98-elongator-smyd3) · NCBI E-utilities 2026
   ▸What this evidence cannot show

   This is an instance of the pattern this project's rules 71 and 77 describe - a query that is accurate as a measurement and false as a conclusion - with a new twist: the failure is not the disease term or the pathway vocabulary but the APPROVED GENE SYMBOL of this project's own index gene. A beat asking 'is SMYD3 connected to KMT2A?' through KMT2A gets 2 records, title-checks both out, and correctly reports zero on-topic records - while a genuine mechanism paper sits one symbol away under MLL. Both off-topic titles were read. Counts taken through the guarded wrapper with querytranslation asserted non-empty. Also measured: SMYD3[tiab] AND lymphoblastic[tiab] = 0 and SMYD3[tiab] AND leukaemia[tiab] = 0 (British spelling) against SMYD3[tiab] AND leukemia[tiab] = 10 and AND myeloid[tiab] = 17 - so SMYD3's leukaemia literature is entirely myeloid and entirely American-spelled, and the B-lineage absence is real.

   ▸Evidence that argues against this (1)

   • The right conclusion is about instruments, not about SMYD3: this project's index gene has an old symbol that carries most of the mechanistic literature, which the septin rule already establishes for other genes and which has now bitten on KMT2A itself. Every SMYD3 query in this entry was therefore run on MLL as well as KMT2A, and this graph should treat 'X AND KMT2A = 0' as uninformative until 'X AND MLL' has been run. Zhang L et al. 2025
9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   IN HUMAN AND MOUSE CHRONIC MYELOID LEUKAEMIA, SMYD3 is over-expressed in leukaemia stem cells and enhances the survival and self-renewal of human primary CML CD34+ cells; loss of SMYD3 blocks leukaemogenesis and impairs the self-renewal and disease-reconstitution abilities of LSCs in mice WITHOUT AFFECTING NORMAL HAEMATOPOIESIS; and SMYD3 stimulates fatty-acid beta-oxidation in LSCs by activating the FABP5/PPARD/CPT1A signalling axis in a methyltransferase-activity-dependent manner, with blockade of CPT1A-mediated FAO reducing human CML LSC function in vitro and depleting LSCs in vivo.

   SMYD3 Activates Fatty Acid β-Oxidation to Promote Self-Renewal of Leukemia Stem Cells. · Zhou M, Wu Z, et al. · Cancer Res 2025
   PMID 40080525DOI
   ▸What this evidence cannot show

   CML, not KMT2A-rearranged leukaemia, and the LSC compartment specifically - stated in the claim per rule 73. Abstract read; full text not read. ALREADY CITED THREE TIMES IN gene-cpt1a (evidence items 6, 11 and 12 of that entry), which also records that the two published CPT1A therapeutic directions oppose each other - inducing CPT1A potentiates antitumour immunity in one system while blocking CPT1A-mediated FAO depletes CML LSCs here. The 'without affecting normal haematopoiesis' clause is the therapeutic-window claim and it is the single most load-bearing sentence for any SMYD3 intervention argument, so it should be read in the paper's own figures before being quoted.

   ▸Evidence that argues against this (2)

   • The DepMap counter is severe and specific: SMYD3 is not a dependency in any of 1,208 lines (pan-essentiality +0.056, i.e. very slightly positive). A rich-medium proliferation screen cannot detect a self-renewal and reconstitution phenotype under metabolic constraint, which is exactly the limitation gene-cpt1a already records for CPT1A itself - so the two results are compatible. But it means the cell-line evidence for SMYD3 as a target is not merely weak, it is absent. Oncofusion knowledge-graph project 2026
   • And the therapeutic direction of the axis is contested downstream: gene-cpt1a records that CPT1A induction sustains a dsRNA-driven type-I interferon response and augments PD-1 blockade in refractory tumours, so 'target the SMYD3 to CPT1A axis' has no determinate sign - context decides, and neither published context is a KMT2A-rearranged leukaemia. Zhang L et al. 2025
10. E11strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Germline deletion of Smyd2 and of the related protein Smyd3 in mice shows both are DISPENSABLE for normal development; ablation of Smyd2 did not affect haematopoiesis but retarded the development of MLL-AF9-promoted leukaemia, Smyd2-deleted leukaemic cells were at a competitive disadvantage in vitro and in vivo, and Smyd2 is directly activated by MYC in MLL-AF9-induced leukaemias.

    Smyd2 is a Myc-regulated gene critical for MLL-AF9 induced leukemogenesis. · Bagislar S, Sabò A, et al. · Oncotarget 2016
    PMID 27655694DOI
    ▸What this evidence cannot show

    Mouse germ-line deletion strains; abstract read, full text not read. This evidence item does double duty and both halves matter. FOR SMYD3 AS A TARGET: a germline Smyd3 knockout with no developmental phenotype is the best available therapeutic-window argument for this protein, and it is stated in one clause of another gene's paper. AGAINST THE SMYD3 STORY'S PRIMACY: the SMYD-family member with mouse genetic evidence in MLL-AF9 leukaemogenesis is SMYD2, and no equivalent Smyd3 test in that model has been reported. SMYD2 (Q9NRG4, 433 aa) and SMYD3 (Q9H7B4, 428 aa) are five residues apart with near-identical protein names, so a swap here would attribute the genetics to the wrong gene and verify-accessions.py could not catch it.

    ▸Evidence that argues against this (1)

    • The paper reports Smyd3's dispensability for development in one clause and does not test Smyd3 in the MLL-AF9 model, so 'Smyd3 does not matter in MLL-AF9 leukaemia' is NOT what it shows - it is an absence of testing, which this project's standing rule says is not a negative result. The SMYD3 result in MLL-AF9 lines came nine years later and used shRNA and a catalytically dead rescue rather than a germline knockout. Zhang L et al. 2025
11. E12established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    EPZ031686, an oxindole sulfonamide identified from a histone-methyltransferase-biased library, is the first orally bioavailable small-molecule SMYD3 inhibitor, with a sulfamide analogue EPZ030456; both have cellular potency sufficient to probe SMYD3 biology in vitro, and EPZ031686 has good bioavailability after oral dosing in mice, making it suitable for in vivo target validation.

    Novel Oxindole Sulfonamides and Sulfamides: EPZ031686, the First Orally Bioavailable Small Molecule SMYD3 Inhibitor. · Mitchell LH, Boriack-Sjodin PA, et al. · ACS Med Chem Lett 2016
    PMID 26985287DOI
    ▸What this evidence cannot show

    A medicinal-chemistry paper reporting TOOL COMPOUNDS, explicitly framed as suitable for target validation rather than as clinical candidates; abstract read, full text not read. No selectivity panel, potency value or cellular IC50 is quoted here because none was taken from the abstract. ChEMBL could not be queried for SMYD3's activity-row count or best reported potency: the activity endpoint returned HTTP 500 on three attempts on 2026-09-22, so the chemical-matter depth for this target is recorded as UNMEASURED rather than as a number - an instrument failure is not a result.

    ▸Evidence that argues against this (1)

    • Ten years after this paper there is still no interventional SMYD3 trial. A first-in-class orally bioavailable tool compound that does not progress in a decade is evidence about the phenotype it produced, and it should be read with the DepMap result: SMYD3 is not a dependency in any of 1,208 cell lines. US National Library of Medicine 2026
12. E13moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    BAY-6035 is a SMYD3 inhibitor discovered by thermal-shift-assay-based high-throughput screening, adding a second chemical series to the SMYD3 probe set alongside the oxindole sulfonamides, BCI-121 and the covalent inhibitor EM127.

    Discovery of the SMYD3 Inhibitor BAY-6035 Using Thermal Shift Assay (TSA)-Based High-Throughput Screening. · Gradl S, Steuber H, et al. · SLAS Discov 2021
    PMID 34154424DOI
    ▸What this evidence cannot show

    A screening-methodology paper; abstract read, full text not read, and no potency or selectivity value is asserted here. The BCI-121 and EM127 names come from PubMed title checking (SMYD3[tiab] AND BCI-121[tiab] = 14 records, SMYD3[tiab] AND EPZ031686[tiab] = 6) and not from a chemistry database, so this item establishes that several independent chemical series exist and asserts nothing about any of their properties.

    ▸Evidence that argues against this (1)

    • Multiple chemical series with no clinical progression is the pattern this project's own rules describe for reading a ChEMBL row count: medicinal chemistry done is not tractability, and a decade of probes with one observational trial is an argument against the target rather than for it. US National Library of Medicine 2026
13. E14moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    A degenerate tetratricopeptide-repeat-like domain in the SMYD3 C-terminal domain mediates physical interaction with HSP90; the C-terminal domain is essential for basal SMYD3 methyltransferase activity and the TPR-like structure is required for HSP90-enhanced activity; and loss of the SMYD3-HSP90 interaction MISLOCALISES SMYD3 WITHIN THE NUCLEUS so that it loses its chromatin association, reducing SMYD3-mediated cell proliferation.

    C-terminal domain of SMYD3 serves as a unique HSP90-regulated motif in oncogenesis. · Brown MA, Foreman K, et al. · Oncotarget 2015
    PMID 25738358DOI
    ▸What this evidence cannot show

    Biochemistry plus cell biology in cancer cell lines; abstract read, full text not read. This is the only SUBNUCLEAR measurement for SMYD3 - a chromatin-bound-versus-soluble-nucleoplasm distinction rather than a compartment name - and it rests on one paper. The mechanism is inferred from mutants that lose HSP90 binding, so 'HSP90 holds SMYD3 on chromatin' is a model; an alternative reading is that the TPR-like mutations destabilise the protein or its activity, which the same paper's basal-activity result shows the CTD is required for.

    ▸Evidence that argues against this (1)

    • HSP90 dependence is a general property of many nuclear clients and is not specific evidence about SMYD3's chromatin function; the paper's own therapeutic framing is that blocking the SMYD3-HSP90 interface would be a lower-toxicity alternative to HSP90 inhibitors, which is a proposal and not a result. No structure of the SMYD3-HSP90 interface is asserted here. Foreman KW et al. 2011
14. E15moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    SMYD3 drives ciliogenesis by direct and indirect regulation of cilia-associated components: it is a novel component of the centriole DISTAL APPENDAGE and is required for centriolar appendage assembly, its loss decreases the percentage of ciliated cells and produces stumpy cilia, it modulates recruitment of the centrosome proteins Cep164, Fbf1, Ninein, Ttbk2 and Cp110 and the trafficking of the intraflagellar transport proteins Ift54 and Ift140, and it binds the promoter regions of C2cd3, Cep164, Ttbk2, Dync2h1 and Cp110.

    SMYD3 Controls Ciliogenesis by Regulating Distinct Centrosomal Proteins and Intraflagellar Transport Trafficking. · Agborbesong E, Zhou JX, et al. · Int J Mol Sci 2024
    PMID 38892227DOI
    ▸What this evidence cannot show

    Abstract read; full text not read. The gene names are given in mouse case in the abstract (C2cd3, Cep164, Ttbk2, Dync2h1, Cp110), so the promoter-binding experiments are most likely in mouse cells; the paper is from a polycystic-kidney-disease group and no leukaemic or haematopoietic cell appears in it. VERIFIES A PREMISE THIS BEAT WAS GIVEN: the coordinator's brief stated that SMYD3 binds the C2cd3 promoter, and it does. It also adds something the brief did not have - SMYD3 is itself a distal-appendage protein, a third subcellular address that no reading of SMYD3 as a chromatin writer predicts. C2CD3 has an entry in this graph as a KMT2A sarcoma partner (gene-c2cd3), where it is a centriole-length gene, so the convergence is real at the level of shared centriolar biology as well as promoter binding.

    ▸Evidence that argues against this (1)

    • The convergence must be recorded with its null stated, as this graph did in cycle 53: SMYD3 binds five cilia-gene promoters in this paper, of which C2cd3 is one, so 'SMYD3 regulates a KMT2A sarcoma partner' is one draw from a five-gene set chosen for ciliary biology and not for any relation to KMT2A. It is a convergence worth one experiment and explicitly not a pattern. Zhou M et al. 2025
15. E16strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.trial_registry

    ClinicalTrials.gov contains exactly ONE registered study naming SMYD3 and it is OBSERVATIONAL: NCT07497789, 'Targeting SMYD3-Driven Metabolic Rewiring and Oxidative Stress Adaptation in Colorectal Liver Metastases', status NOT_YET_RECRUITING, enrolment 156, start date 2026-05-10, sponsored by Azienda Ospedaliera Specializzata in Gastroenterologia Saverio de Bellis - so no interventional trial of any SMYD3 inhibitor has ever been registered.

    ClinicalTrials.gov intervention search, run 2026-09-22 · US National Library of Medicine · ClinicalTrials.gov 2026
    ▸What this evidence cannot show

    Run 2026-09-22 as an intervention search on 'SMYD3 OR EPZ031686 OR BAY-6035' with count_total, returning total = 1. POSITIVE CONTROL THROUGH THE IDENTICAL TOOL AND QUERY SHAPE: 'revumenib OR ziftomenib' returns total = 43, so the instrument is working (rule 67). The name list is the bound and is stated as such - EPZ030456, BCI-121, EM127, GSK2807 and any unpublished clinical code were NOT searched, and a compound in trials under a company code would be invisible to this query (this project's rule 72). An intervention-field search also cannot see a trial that mentions SMYD3 only as a biomarker in its outcome measures.

    ▸Evidence that argues against this (1)

    • A registry absence is bibliographic in the same way a PubMed absence is: a first-in-human study under an undisclosed code, or a trial using SMYD3 only as a stratification biomarker, would not appear. The name list, not the query, is the limit - and the correct reading is 'no trial names SMYD3 or its two best-known probes as an intervention', which is weaker than 'no SMYD3 inhibitor has ever been given to a patient'. Mitchell LH et al. 2016
16. E10weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.dataset

    SMYD3 shows no KMT2A-selective dependency in this project's own DepMap scan: mean Chronos -0.0138 in 11 KMT2A-rearranged lines against 0.0456 in 43 other leukaemia lines, difference -0.0595, at a pan-essentiality of 0.0564 across 1208 lines and selectivity rank 2811 of 17,728.

    DepMap CRISPR (Chronos) gene-effect scan of 11 KMT2A-rearranged versus 43 other leukaemia lines (this project's own genome-wide scan, analyses/depmap-kmt2a-selective-scan-full) · Oncofusion knowledge-graph project · analyses/depmap-kmt2a-selective-scan-full/results.json 2026
    ▸What this evidence cannot show

    Joined 2026-09-22 from analyses/depmap-kmt2a-selective-scan-full/results.json (generated 2026-09-19T10:12:06Z). NOT re-fetched and NOT independent of that scan: every caveat it carries applies, including that the 11 KMT2A-rearranged lines come from nine individuals and that the arm is myeloid-enriched, so a difference there mixes fusion status with lineage. The scan records a generation timestamp and NOT a DepMap RELEASE, which is this project's rule-83 exposure carried forward rather than fixed here. Read the numbers by this project's own DepMap rules: a rank is not an effect size, a difference is not a dependency, the anti-selective half is 50 percent of any gene set by construction with this graph's base rate at 45.3 percent, and a gene depleted this deeply across all lines is FLOOR-COMPRESSED - there is almost no dynamic range left in which selectivity could appear, so a small arm-to-arm difference is uninterpretable rather than small. The screen is also leukaemia lines only.

    ▸Evidence that argues against this (3)

    • A difference of -0.0595 at a pan-essentiality of 0.056 across 1208 lines is floor-compressed and uninterpretable, not small: a gene required in nearly every line has no dynamic range left in which KMT2A selectivity could show. Reporting the rank would be reporting the null. Oncofusion knowledge-graph project 2026
    • THIS ONE IS NOT FLOOR COMPRESSION AND IT IS THE STRONGER READING: SMYD3's pan-essentiality is +0.056, i.e. very slightly POSITIVE across 1,208 lines, and BOTH arms sit at zero (-0.014 in KMT2A-rearranged, +0.046 in other leukaemia). A gene that is positive in both arms is not a dependency at all, so the -0.060 difference and the rank of 2,811 are ranking noise between two nulls - this project's rule 84 in its sharpened form. The paralogue is the same: SMYD2 pan-essentiality +0.036. Neither SMYD protein is required by any cell line in the screen. Oncofusion knowledge-graph project 2026
    • The two leukaemia results are the counter-argument to that and both are conditional in ways a proliferation screen cannot see: the CML result is a self-renewal and disease-reconstitution phenotype under metabolic constraint in a rich-medium screen's blind spot - the exact limitation gene-cpt1a already records for CPT1A - and the AML result is a combination with DOT1L inhibition rather than a single-agent dependency. Zhou M et al. 2025

• regulates →HOXA9 — required for survival of established MLL-rearranged leukemia, and dispensable for starting it

  SMYD3-H3K4me3-HOXA9 feed-forward loop in MLL-AF9 AML: SMYD3 mediates H3K4me3 at the HOXA9 promoter and activates its transcription, HOXA9 binds the SMYD3 promoter in return; full-text reading and counter-evidence live in gene-hoxa9 E8

• related to →H3K4me3 - the mark KMT2A's own SET domain writes, that the fusion cannot write, that an interface in this graph already binds, and that switches PRC2 off

  the mark SMYD3 is usually credited with (H3K4me2/me3, not me1) and the one functionally tested in MLL-AF9 AML - but two other histone marks are independently claimed, H4K5me and an H4K20me3 preference

• regulates →CPT1A - the outer-mitochondrial-membrane fatty-acid gatekeeper that is also a lysine succinyltransferase, fused to KMT2A exon 3 in the one case that is tether WITHOUT address

  SMYD3 activates the FABP5/PPARD/CPT1A fatty-acid-beta-oxidation axis in CML leukaemia stem cells in a methyltransferase-activity-dependent manner; already cited three times in gene-cpt1a, which records that the two CPT1A therapeutic directions oppose each other

• regulates →C2CD3 - the centriole-elongation regulator fused to KMT2A exon 8 in an MPNST, predicted OUT OF FRAME, and the partner whose RefSeq transcript is 390 residues shorter than UniProt's canonical

  SMYD3 binds the C2cd3 promoter and is itself a centriole distal-appendage protein; recorded as a convergence WITH ITS NULL STATED - C2cd3 is one of five cilia promoters that paper reports - and explicitly not as a pattern

• related to →DOT1L — the target where excellent chemistry met an indirect interaction

  combined disruption of the SMYD3-HOXA9 loop and DOT1L inhibition enhanced anti-leukaemia activity in MLL-AF9 AML in vitro and in vivo; SMYD3 activates HOXA9 independently of DOT1L

• related to →KMT2A::MLLT3 (MLL-AF9), t(9;11)(p21.3;q23.3)

  the axis is MLL-AF9 / KMT2A::MLLT3, not the index fusion: SMYD3 and HOXA9 induction is GREATER in MOLM-13, THP-1 and NOMO-1 than in MV4-11, which carries KMT2A::AFF1

• related to →The differentiation block — what it actually is, operationally, in KMT2A::AFF1 leukaemia

  the SMYD3-H3K4me3-HOXA9 loop exacerbates proliferation and blocks differentiation in MLL-AF9 AML lines and primary patient cells

• related to →The soluble-nucleoplasm versus chromatin-bound axis - the one localisation readout that keeps a size dimension, and nobody has run it on a KMT2A fusion

  SMYD3 chromatin association is HSP90-dependent: losing the SMYD3-HSP90 interaction mislocalises SMYD3 WITHIN the nucleus and it loses chromatin association - a chromatin-bound-versus-soluble distinction, the only subnuclear measurement for this protein

• related to →KRAS - the commonest co-mutation in a disease that has almost none, and it is usually in a minority of the cells

  SMYD3 methylates MAP3K2 at K260, blocking PP2A binding and potentiating Ras/Raf/MEK/ERK; SMYD3 catalytic activity is required for Ras-driven pancreatic and lung tumour development in mice and synergises with a MEK inhibitor

• related to →KMT2A / MLL1 — the gene that breaks

  IDENTIFIER TRAP ON THIS PROJECT OWN INDEX GENE: SMYD3[tiab] AND KMT2A[tiab] = 2 and BOTH are off topic, while AND MLL[tiab] = 5 and AND HOXA9[tiab] = 1 carry the mechanism. Treat X AND KMT2A = 0 as uninformative until X AND MLL has been run

• related to →Pinometostat (EPZ-5676) — the DOT1L inhibitor, and the most instructive failure in this field

  the only supported SMYD3 combination in a KMT2A fusion is with DOT1L inhibition, so a SMYD3 combination rationale inherits pinometostat clinical history

• related to →KMT2A-rearranged acute myeloid leukaemia in children and infants

  SMYD3 is over-expressed in AML and predicts poor prognosis; the HOXA9 loop was shown in MLL-AF9 AML lines and primary AML patient fractions