SETD1B - the COMPASS paralogue round 4 deferred for want of a leukaemia paper, which had already been published: its SET domain sets H3K4me3 breadth and MYC in MLL-rearranged AML

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.

• Complete Setd1b loss in normal adult mouse haematopoiesis causes multilineage dysplasia and is lethal. Any SETD1B-directed agent must beat that, and whole-protein degradation is therefore the wrong modality from the outset.
• The requirement is catalytic and domain-localised, so a catalytic-site inhibitor is the natural modality - but SET-domain inhibitors in this family have repeatedly failed to achieve paralogue selectivity, and there are six paralogues.
• SETD1B carries its own WIN motif at 1745-1750, so a WDR5 WIN-site inhibitor would hit it as collateral damage - relevant because this graph already holds a WIN-site target dossier.
• If MYC rescue is complete, the therapeutic index is set by MYC dependence rather than by SETD1B, and MYC-amplified or MYC-translocated relapse would be an obvious escape route.
• The dependency signal is myeloid-only in the lineage split, and this project's index disease is infant B-ALL. A programme reading rank 1,311 as an infant-ALL lead would be reading the wrong half of the data.

1. Does the SETD1B SET-domain requirement hold in KMT2A::AFF1 B-ALL, or only in AML with a cooperating FLT3/RAS lesion? The lineage split says the dependency signal is myeloid-only, and the published tiling was done in AML, so this is genuinely open and directly testable in SEM or RS4;11.
2. Is the requirement a MYC requirement wearing a chromatin costume? Exogenous MYC rescues the growth defect, which raises the possibility that SETD1B matters here only insofar as it sustains MYC - a much less interesting conclusion that the published rescue is consistent with.
3. Why does disrupting KDM5C rescue? If the phenotype is a net H3K4me3-breadth deficit, then blocking the eraser should help - which makes KDM5C, not SETD1B, the more tractable node and nobody appears to have pursued it.
4. SETD1B and SETD1A give opposite answers on catalytic dependence in AML. Is that a real paralogue divergence or a difference between the assays used, and does a tiling screen of SETD1A in the same system reproduce the FLOS result?

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

   A CRISPR-tiling screen against known H3K4 methylation modifiers in an MLL-rearranged AML model showed that disrupting the SETD1B catalytic SET domain depletes FLT3-ITD- or NrasG12D-expressing AML cells and downregulates gene expression particularly in the MYC pathway; SETD1B SET-domain loss significantly decreases H3K4me3 breadth; and exogenous MYC expression or disruption of the H3K4 demethylase KDM5C significantly restores growth defects in SETD1B SET-domain-mutant cells.

   Regulation of H3K4me3 breadth and MYC expression by the SETD1B catalytic domain in MLL-rearranged leukemia · Izumi S, Ohtani K, Matsumoto M, Shibata S, Rahmutulla B, Fukuyo M, Nishimoto M, Miyagawa H, Sakaida E, Yokote K, Kitabayashi I, Araki K, Kaneda A, Hoshii T · Leukemia 39(7):1627-1639 2025
   PMID 40341256DOI
   ▸What this evidence cannot show

   Abstract fetched from PubMed and read on 2026-09-19; full text not read, and every identifier copied from that fetched record. **The cells carry cooperating FLT3-ITD or NrasG12D lesions**, so the requirement is shown in that context and not in a bare MLL-rearranged background. AML, not the infant B-ALL this project indexes on. 'H3K4me3 breadth' is a derived metric sensitive to peak-calling choices that the abstract does not specify. The two rescues (MYC, KDM5C) are the strongest element and were read only as abstract statements.

   ▸Evidence that argues against this (1)

   • The paralogue SETD1A presents the opposite picture in AML - its dependency is catalytically independent, with the SET domain dispensable and a separate FLOS domain recruiting cyclin K - so 'SET1-family member, therefore catalytic requirement' is not a safe generalisation, and the SETD1B result should not be extended to the family. Schmidt K et al. 2018
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   Conditional mutagenesis of Setd1b in adult mice, by both ubiquitous and haematopoietic-specific deletion, caused peripheral thrombocytopenia and lymphocytopenia, multilineage dysplasia, myeloid-biased extramedullary haematopoiesis in the spleen and lethality; transplantation and expression profiling showed Setd1b is autonomously required in haematopoietic lineages, where it regulates key lineage-specification components including Cebpa, Gata1 and Klf1.

   The H3K4 methyltransferase Setd1b is essential for hematopoietic stem and progenitor cell homeostasis in mice · Schmidt K, Zhang Q, Tasdogan A, Petzold A, Dahl A, Arneth BM, Slany R, Fehling HJ, Kranz A, Stewart AF, Anastassiadis K · eLife 7:e27157 2018
   PMID 29916805DOI
   ▸What this evidence cannot show

   Abstract fetched from PubMed and read on 2026-09-19; full text not read, identifiers copied from that fetched record. **Mouse, normal haematopoiesis, no leukaemia and no fusion anywhere in this work.** It bounds the toxicity of SETD1B loss in normal blood and says nothing about a therapeutic window in leukaemia. Adult conditional deletion, so developmental requirements are not addressed.

   ▸Evidence that argues against this (1)

   • This result is the strongest argument AGAINST SETD1B as a target: complete loss in normal adult haematopoiesis is lethal with multilineage dysplasia, which is exactly the profile a paediatric programme cannot accept. The leukaemia result concerns SET-domain disruption rather than whole-gene loss, so the two are not strictly comparable, and that difference is where any window would have to live. Izumi S et al. 2025
3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   UniProtKB Q9UPS6 (SET1B_HUMAN, Histone-lysine N-methyltransferase SETD1B) is 1,966 aa with synonyms KIAA1076, KMT2G and SET1B; it is annotated to nucleus, nuclear speck, chromosome and cytoplasm; it has four PDB cross-references (3UVO, 4ES0, 8ILZ, 8WXV); and it carries an RRM domain at 93-181, a WDR82-interaction region at 68-98, a coiled coil at 1173-1204, a WDR5-interaction (WIN) motif at 1745-1750, an RxxxRR motif at 1798-1803, a SET domain at 1827-1944 and a post-SET domain at 1950-1966.

   UniProtKB Q9UPS6 (SET1B_HUMAN), Histone-lysine N-methyltransferase SETD1B · UniProt Knowledgebase, reviewed (Swiss-Prot) 2026
   ▸What this evidence cannot show

   Record fetched 2026-09-19 via UniProt REST with gene_exact:SETD1B AND organism_id:9606 AND reviewed:true; **primary gene name read back as SETD1B and length 1,966 confirmed from the same fetched record**, per the standing accession rule. One reviewed hit. All seven boundaries copied from the fetched feature table; none estimated.

   ▸Evidence that argues against this (1)

   • SETD1B carries its own WIN motif at 1745-1750, so it competes for the same WDR5 pocket that this graph's WIN-motif target dossier is built around and that the NSL scaffolds also engage. A WIN-site inhibitor would not be SETD1B-sparing, which matters given that complete Setd1b loss is lethal in mouse haematopoiesis. Schmidt K et al. 2018
4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

   In the genome-wide KMT2A-rearranged selectivity scan of 17,728 genes, SETD1B has a pan-essentiality of -0.2403 across 1,208 lines and a KMT2A-rearranged difference of -0.0931 (KMT2A-rearranged mean -0.4473 against other-leukaemia mean -0.3542), ranking 1,311 (7.395th percentile). Its paralogue SETD1A sits at pan -0.91.

   Genome-wide KMT2A-rearranged selectivity scan, 17,728 genes (DepMap Chronos via Open Targets) · Broad Institute DepMap / Open Targets Platform; analyses/depmap-kmt2a-selective-scan-full 2026
   ▸What this evidence cannot show

   Row read out of analyses/depmap-kmt2a-selective-scan-full/results.json on 2026-09-19 in a scripted pass. Eleven KMT2A-rearranged against 43 other leukaemia lines, AML-heavy, none infant-derived. **A difference of -0.093 is inside noise**, and the 7th-percentile rank compresses hard near the distribution centre. This entry exists on the strength of the literature, not this number, and says so.

   ▸Evidence that argues against this (1)

   • A whole-gene CRISPR knockout is the wrong readout for a requirement demonstrated by SET-domain tiling. If the cells tolerate loss of the protein but not loss of its catalytic domain - which is what a separation-of-function result usually means - a knockout screen would show exactly the weak signal seen here. Izumi S et al. 2025
5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

   On the lineage-split run for this beat, SETD1B's requirement is myeloid-restricted and small: KMT2A-rearranged myeloid z -0.65, B-lineage z +0.18, split gap -0.83, OCI-AML3 z +1.91, U-937 z +0.90, on a pan-essentiality of -0.241. The run classifies it 'weak or ambiguous' and does not flag it as a lineage artefact.

   Interactome round 5 panel, lineage-split DepMap dependency (FACT, NSL, MSL, COMPASS remainder, Mediator) · Broad Institute DepMap / Open Targets Platform; analyses/depmap-interactome5-lineage-split 2026
   ▸What this evidence cannot show

   Rows read from analyses/depmap-interactome5-lineage-split/results.json on 2026-09-19. Six myeloid and four B-lineage index lines, so a single line moves a mean; the split is by lineage LABEL, not expression signature. **Reporting both halves is mandatory here**: the single-number genome-wide rank of 1,311 conceals that the B-lineage half is positive.

   ▸Evidence that argues against this (1)

   • SETD1B regulating Cebpa in normal haematopoiesis gives a mechanistic reason to expect a myeloid-weighted dependency independent of any KMT2A fusion, which is the same confound that makes CEBPA itself rank 25 genome-wide with no dependency in KMT2A-rearranged B-ALL. The myeloid-only signal here should be read in that light. Schmidt K et al. 2018
6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

   As of 2026-09-19, '(MLL OR KMT2A) AND (SETD1B OR KMT2G OR KIAA1076 OR Set1B)' returns 27 records and '(SETD1B OR KMT2G) AND (leukemia OR leukaemia)' returns 21. The round 4 enumeration of this beat ran '(MLL OR KMT2A) AND (SETD1B OR KMT2G OR KIAA1076)' and recorded 17; adding the spelling 'Set1B' raises the count to 27 and the MLL-rearranged AML paper is reachable in the larger set.

   PubMed E-utilities esearch record counts for the COMPASS remainder (SETD1B, CXXC1) against KMT2A, MLL and leukaemia, run for interactome round 5 · NCBI PubMed, queried 2026-09-19 2026
   ▸What this evidence cannot show

   Counts from the esearch `count` field, queried 2026-09-19. The 27-record set was scanned by title, not read; it contains a substantial neurodevelopmental and structural-biology literature alongside the haematology. The round 4 count of 17 is quoted from that beat's own recorded query string in `ana-interactome4-enumeration` and was not re-run here in its original form, so the 17-versus-27 comparison is between this session's run and that entry's record.

   ▸Evidence that argues against this (1)

   • The difference between the two queries may partly reflect PubMed term mapping rather than a clean alias effect, since PubMed expands gene symbols to supplementary-concept records and two queries that look equivalent can differ for reasons unrelated to spelling. The robust claim is that the relevant paper is reachable from one query and was not recorded from the other; the mechanism is inferred. Izumi S et al. 2025

• related to →SETD1A — a leukemia dependency whose enzyme activity turns out to be dispensable

  Paralogues with opposite answers on catalytic dependence in AML: SETD1B's SET domain is required, SETD1A's is dispensable.

• binds →WDR5 — the partner that is not a partner: it binds the half of KMT2A the fusion throws away

  SETD1B carries its own WIN motif at residues 1745-1750, so it competes for the WDR5 pocket the graph's WIN-site dossier targets.

• part of →CXXC1 (CFP1) - a CpG-reading COMPASS subunit that binds MLL1 and recruits it to HoxA7, and whose CXXC domain cannot substitute for KMT2A's in a fusion

  Both are components of SET1-family COMPASS complexes; CXXC1 targets SETD1A/B to unmethylated CpG islands.

• regulates →CEBPA — ranks 25th on the KMT2A scan and sits at the median of random genes

  Setd1b regulates Cebpa among key lineage-specification components in mouse haematopoiesis - a mechanistic route to a myeloid-weighted dependency independent of any fusion.

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

  SET-domain disruption depletes MLL-rearranged AML cells carrying FLT3-ITD or NrasG12D and narrows H3K4me3 breadth.

• ← related toInteractome round 5: 42 queries, 13 entries written, 9 candidates rejected with counts, and four alias failures of four different kinds

  Written this cycle after the round 4 deferral was found to rest on an incomplete symbol set.

• ← related toKMT2B (MLL2/MLL4) - the other protein menin binds, and the one this graph never wrote

  SET1/MLL family; menin is absent from the SET1A/B complexes