PHGDH (O43175) - the largest menin-SELECTIVE fall in the whole serine panel, and the least essential gene in it

This is real experimental data — atoms placed by crystallography or cryo-EM, deposited by the people who did the experiment.

• **Germline biallelic PHGDH loss causes Neu-Laxova syndrome 1: lethal, with intrauterine growth restriction, microcephaly, hypoplastic lungs and lissencephaly.** The milder allelic form causes congenital microcephaly, psychomotor retardation and seizures. The target population here is INFANTS, in whom neurodevelopment and growth are the clinical priorities immediately behind survival, and a PHGDH inhibitor would be given during the period the germline phenotype maps onto. This is the strongest essential-function objection in this beat and it is not mitigated by the flat DepMap profile - a CRISPR screen in a cell line cannot see a developmental phenotype.
• **A dependency figure must not become a target nomination, and here the dependency figure actively refuses it.** PHGDH falls ~10-fold under a menin inhibitor and 1.1% of 1,258 DepMap lines need it. A large drug-induced fold change in a non-essential gene is most parsimoniously a MARKER OF RESPONSE, not a vulnerability. If PHGDH's fall is a readout of fusion-programme collapse, then adding a PHGDH inhibitor to a menin inhibitor is adding a second hit to a node the first drug has already removed, which is the opposite of synergy.
• **No PHGDH inhibitor has ever been in a human trial**, and both the compounds this literature runs on are explicitly tool compounds - CBR-5884 is a noncompetitive oligomerisation disruptor with time-dependent onset from a 2016 screen, NCT-503 is a screen hit with an off-mechanism effect on exogenous one-carbon units. Nothing in this entry supports naming PHGDH as a drug target in a paediatric protocol; it supports naming it as a mechanism to understand.
• **Serine and glycine are required for nucleotide synthesis, glutathione and one-carbon units in every proliferating tissue**, including the marrow this patient needs to recover and the gut epithelium chemotherapy has already injured. The one clinical serine-restriction protocol on the registry is a phase 1/2 diet trial in adults with rectal cancer (NCT07811076); there is no analogue for an infant, in whom growth is the endpoint.
• **Serine restriction acts partly by causing ROS and DNA damage** (PMID 32766858). Adding a genotoxic pressure to a leukaemia already treated with anthracyclines, etoposide and cytarabine is a route to therapy-related second malignancy, which is already a documented concern in this population.
• **If the serine collapse is menin-selective rather than a generic fusion-programme readout, that is bad news for a DOT1L inhibitor and good news for nothing.** The clinically relevant reading of the asymmetry is that pinometostat would NOT reproduce this arm of a menin inhibitor's effect, not that PHGDH is a target.

1. Does H3K4me3 at the PHGDH promoter fall under VTP-50469 in RS4;11, and is menin or the KMT2A fusion resident there? This is one ChIP in cells two laboratories already grow, and it is the direct test of whether the menin-selective transcript margin measured here reflects the chromatin input PMID 29672864 demonstrated in Ewing sarcoma.
2. Is the menin-versus-DOT1L margin a dose artefact? A dose-response of both compounds on PHGDH and CHAC1, with the DOT1L dose titrated to match VTP-50469 on MEIS1, would settle it. This is the single most serious alternative explanation for the beat's main finding and it applies to every gene in it.
3. Is PHGDH a dependency in a KMT2A::AFF1 model at all? DepMap says 1.1% of 1,258 lines need it, in serine-replete medium. The experiment that matters is a PHGDH knockout or NCT-503 exposure in RS4;11 and SEM in serine-free and serine-replete medium, with and without a menin inhibitor - which also tests whether menin inhibition creates a serine dependency rather than exploiting one.
4. Does PHGDH protein fall, and does serine flux change, in KMT2A-rearranged cells under a menin inhibitor? Every number in this entry is transcript. PMID 29672864 measured 13C flux in Ewing sarcoma; nobody has measured it in this disease.
5. Is PHGDH expression heterogeneous across KMT2A-rearranged patient samples, and does it track fusion partner or lineage? The transcript falls hard in RS4;11, SEM, MV4;11 and MOLM13 under drug, but the BASELINE has never been compared across primary infant samples, and PHGDH-high versus PHGDH-low is the split that predicts inhibitor response everywhere else it has been measured.

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

   In RS4;11 (KMT2A::AFF1) at day 7 on a pre-registered 41-symbol panel, PHGDH falls 3.24 log2 under VTP-50469 at 330 nM from a DMSO baseline of 7.61 log2 CPM and only 1.52 log2 under EPZ5676 at 1 uM in the same experiment, a VTP-minus-EPZ margin of -1.72 that is the largest in the serine/one-carbon panel and exceeds the canonical menin target PBX3 (-1.02); across the panel the serine arm's median margin is -0.55 (n=9) against the ISR-output arm's -0.024 (n=4, range -0.085 to +0.095), Mann-Whitney two-sided p = 0.011, while the serine arm is not separable from the menin-target positive control (median -1.02, p = 0.106).

   analyses/q109-is-the-serine-collapse-downstream-of-atf4-or-parallel - panel-preregistration.json written to disk before any expression value was read and loaded by run.py; GSE127506 + GSE127508 3-prime-end RNA-seq count matrix, RS4;11 and MOLM13, day 2 and day 7, VTP-50469 330 nM and EPZ5676 1 uM against same-day DMSO, every fold change reported with its own DMSO baseline and a floor-collision test · this repository 2026
   ▸What this evidence cannot show

   One cell line, one timepoint, two single doses and no dose-response, so a saturation artefact cannot be excluded: EPZ5676 at 1 uM may simply be further from saturation on this locus than on CHAC1. n=2-3 per arm; no multiple-testing correction is applied and none is claimed. The Mann-Whitney p-values rest on n=9 against n=4 and are reported for completeness - the per-gene table is the result, not the p-value. The 3-prime-end assay measures transcript abundance only. No medium amino-acid composition is recorded in the deposited metadata, which is a live alternative explanation for the whole amino-acid-response arm. Day 7 of continuous exposure to an effective drug is late enough that arrest is a competing explanation for any coordinated biosynthetic fall.

   ▸Evidence that argues against this (1)

   • The menin-target positive control separates the two compounds by 1.0-3.8 log2, so a compound with a weaker effect on the fusion programme will produce a smaller effect on anything downstream of it, PHGDH included - which is exactly what a single-input model with different transmission efficiency predicts. The comparison that would settle this is a dose of EPZ5676 chosen to match VTP-50469 on MEIS1, and it does not exist in these data. this repository 2026
2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   The same menin-versus-DOT1L asymmetry reproduces in MV4;11 (KMT2A::AFF1, myeloid/biphenotypic) from an independent laboratory and a different series: PHGDH -2.77 under VTP-50469 at 500 nM against -0.88 under the DOT1L inhibitor SGC0946 at 5 uM, a margin of -1.89, while CHAC1 is -1.17 against -1.18 (margin +0.02); across the panel the serine arm's median menin-minus-DOT1L margin is -1.05 (n=7) against the ISR-output arm's +0.02 (n=2).

   DOT1L provides transcriptional memory through PRC1.1 antagonism. · Neville D, Ferguson DT, Heikamp EB, et al. · Nature Cell Biology 2026
   PMID 41634341DOI
   ▸What this evidence cannot show

   The comparator here is SGC0946 at 5 uM, a DIFFERENT DOT1L inhibitor at a different dose from EPZ5676 at 1 uM, so this arm tests the DIRECTION of the asymmetry and not its magnitude. Housekeeping is not flat in this matrix - ACTB is -0.73 in the Safe3 menin arm - so library-size CPM is inflating every repression here and the absolute numbers should be read with the housekeeping-adjusted column in the cycle-56 analysis. The ISR-output comparison rests on only two usable genes because DDIT3 (baseline 1.93 log2 CPM), CHAC1 (1.90) and TRIB3 (3.00) are near the detection floor in MV4;11. MV4;11 is myeloid/biphenotypic where RS4;11 and SEM are B-lymphoid, so this crosses a lineage boundary.

   ▸Evidence that argues against this (1)

   • In this matrix the housekeeping set itself shows a menin-minus-DOT1L margin (ACTB -0.42, TBP -0.47, RPL13A -0.73), so part of the serine arm's -1.05 median margin is a global normalisation artefact rather than gene-specific selectivity. Subtracting the housekeeping margin would leave PHGDH's margin around -1.4, still well outside the ISR arm's, but the correction is not negligible and the ISR arm's near-zero margin is what makes the contrast survive it. this repository 2026
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   In Ewing sarcoma, Bru-seq after the menin-MLL interaction inhibitor MI-503 showed the serine biosynthetic pathway to be the pathway most significantly affected; baseline expression of PHGDH, PSAT1 and PSPH and metabolic flux through the pathway are high; inhibition of PHGDH reduced cell proliferation, viability and tumour growth in vivo; and inhibition of menin diminished expression of SSP genes, REDUCED H3K4me3 ENRICHMENT AT THE PHGDH PROMOTER, and completely abrogated de novo serine and glycine biosynthesis as shown by 13C-glucose metabolic tracing.

   Menin regulates the serine biosynthetic pathway in Ewing sarcoma. · Svoboda LK, Teh SSK, Sud S, et al., Lawlor ER · J Pathol 2018
   PMID 29672864DOI
   ▸What this evidence cannot show

   Abstract read this session; full text not retrieved, so the magnitude of the H3K4me3 change, the antibody, the promoter coordinates and the number of replicates were not checked. **The system is Ewing sarcoma, where KMT2A/MLL1 and menin are overexpressed WILD-TYPE oncogenes and there is no KMT2A fusion**, so this is not the index disease and not even the index lesion class. The PHGDH-inhibition arm is a tool compound in that system. No equivalent H3K4me3 ChIP at the PHGDH promoter exists in any KMT2A-rearranged cell, which is this claim's largest gap and is one ChIP away from closing.

   ▸Evidence that argues against this (1)

   • H3K4me3 loss at a promoter under a menin inhibitor is expected wherever the menin-MLL complex is resident and is not by itself evidence of a direct or specific input: the same paper reports early and widespread reprogramming of metabolic processes, so PHGDH may be one locus in a broad chromatin response rather than a selected target. And because menin and MLL1 PROTEIN are lost under MI-503 in this system, a chromatin change downstream of protein loss cannot be separated from an interface-specific effect. Svoboda LK et al. 2018
4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   In Ewing sarcoma ATF4 levels are high and ATF4 modulates transcription of core serine synthesis pathway genes; inhibition of either EWS-FLI1 or menin leads to loss of ATF4 with diminished expression of SSP transcripts AND proteins; an EWS-FLI1 binding site at the ATF4 promoter was identified and validated; menin-dependent regulation of ATF4 is mediated by transcriptional AND post-transcriptional mechanisms; and menin inhibition similarly leads to loss of ATF4 and the ATF4-dependent transcriptional signature in MLL-rearranged B-cell acute lymphoblastic leukemia.

   EWS-FLI1 and Menin Converge to Regulate ATF4 Activity in Ewing Sarcoma. · Jimenez JA, Apfelbaum AA, Hawkins AG, et al., Lawlor ER · Mol Cancer Res 2021
   PMID 33741715DOI
   ▸What this evidence cannot show

   Abstract read this session; full text not retrieved. The MLL-rearranged B-ALL extension is a single sentence in the abstract and the fusion partners of those lines are not stated there, so whether any of them is KMT2A::AFF1 is unknown and is a live open question in `gene-atf4`. The paper attributes SSP loss to ATF4 loss; it does not exclude a parallel input, and its own companion paper reports the H3K4me3 change at the PHGDH promoter [E3]. Cited in 7 entries of this graph already - read those before adding to this one.

   ▸Evidence that argues against this (1)

   • This graph's own measurement complicates the mechanism: ATF4's OWN transcript does not move under either a menin or a DOT1L inhibitor in RS4;11 (max -0.86 log2 while CHAC1 is -5.11 in the same samples, on a transcript sitting at the array ceiling), and the whole eIF2alpha kinase module is flat. So in leukaemia the ATF4 step is an activity change that a transcript assay cannot see - which means an 'ATF4-dependent' attribution in this disease currently rests on target-gene behaviour, and the target genes do not all behave alike. this repository 2026
5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   PHGDH is not a dependency in any lineage DepMap screens: gene effect mean -0.0795, median -0.0725, 1.1% of 1,258 screens below -0.5, and a leukaemia mean of -0.0685 over 57 screens, fetched with every one of the 29 tissue blocks flattened and the resolved Ensembl id's approvedSymbol read back as PHGDH.

   Open Targets Platform GraphQL target.depMapEssentiality, all 29 tissue blocks flattened, panel-maximum payload guard 1,258 screens PASS; analyses/q109-serine-one-carbon-depmap-dependency/results.json · Open Targets / DepMap 2026
   ▸What this evidence cannot show

   Rule 83: this payload carries NO release identifier, so `depmap_release` is recorded as null and the absolute values may drift between quarterly releases; the cross-gene CONTRAST is the durable part. The 57 leukaemia screens are identified by the payload's own diseaseFromSource string matched against four substrings, not by a curated lineage table, and they are NOT split by KMT2A status here - this is a lineage figure, not a fusion figure. A CRISPR knockout in serine-replete medium is close to the least sensitive possible assay for a biosynthetic enzyme whose product is in the medium, so this null is partly a property of culture conditions.

   ▸Evidence that argues against this (1)

   • Standard DepMap medium contains serine, so a knockout of the enzyme that MAKES serine is measured under conditions that mask the phenotype. PMID 21760589 found PHGDH suppression to be strongly antiproliferative specifically in PHGDH-HIGH lines and not in others, and PMID 27110680 reports that genetic loss of PHGDH is toxic to PHGDH-overexpressing lines EVEN WITH exogenous serine present - so a flat pan-lineage mean is compatible with a real dependency in a subset defined by expression, which is not how the DepMap summary is indexed. Possemato R et al. 2011
6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   UniProt reviewed human record O43175 (SERA_HUMAN) has primary gene name PHGDH, length 533 aa, catalyses the reversible oxidation of 3-phospho-D-glycerate to 3-phosphonooxypyruvate as the first step of phosphorylated L-serine biosynthesis, carries 21 PDB cross-references, is positively regulated by SP1 and NF-Y, and records TWO diseases: phosphoglycerate dehydrogenase deficiency (congenital microcephaly, psychomotor retardation, seizures) and Neu-Laxova syndrome 1, a lethal autosomal recessive multiple-malformation syndrome with ichthyosis, marked intrauterine growth restriction, microcephaly, limb deformities, hypoplastic lungs and CNS anomalies including lissencephaly. The record carries NO SUBCELLULAR LOCATION comment of any kind.

   UniProtKB entry O43175 (SERA_HUMAN), D-3-phosphoglycerate dehydrogenase · UniProt Consortium · UniProt 2026
   ▸What this evidence cannot show

   Resolved 2026-09-22 with gene_exact:PHGDH AND organism_id:9606 AND reviewed:true; exactly one reviewed row, gene_primary and length read back. The absent SUBCELLULAR LOCATION comment is a fact about the database, not about the protein: it means UniProt has curated no localisation statement, coded or uncoded, and this beat did not open a primary imaging paper to fill it. The disease descriptions are curated summaries, not a cohort; the allelic series linking mild deficiency to lethal Neu-Laxova is a genotype-phenotype claim the record asserts and this entry has not independently checked.

   ▸Evidence that argues against this (1)

   • A germline null phenotype is a poor guide to the toxicity of partial pharmacological inhibition in a post-natal patient: Neu-Laxova is an intrauterine phenotype and phosphoglycerate dehydrogenase deficiency is reported as at least partly treatable by amino-acid supplementation, which implies the pathway's loss is survivable when serine is supplied exogenously. That cuts both ways for a leukaemia patient on parenteral nutrition. de Koning TJ 2006
7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.trial_registry

   There is no interventional clinical trial of any PHGDH inhibitor: ClinicalTrials.gov returns 5 studies for the term PHGDH and none is an interventional trial of a PHGDH inhibitor, the term NCT-503 returns 0 and CBR-5884 returns 0; the entire clinical footprint of serine-pathway targeting is one not-yet-recruiting phase 1/2 randomised trial of a serine/glycine-restricted diet with chemoradiotherapy and immunotherapy in locally advanced rectal cancer, NCT07811076, 140 estimated participants, estimated start 2026-08-31.

   A Randomized Controlled Trial of Serine/Glycine-Restricted Diet Combined With Chemoradiotherapy and Immunotherapy for Neoadjuvant Treatment of Rectal Cancer · ClinicalTrials.gov 2026
   NCT07811076
   ▸What this evidence cannot show

   Queried through the raw ClinicalTrials.gov v2 API on 2026-09-22 with free-text terms; a registry free-text search can miss a trial that names only a company compound code, and an unregistered or investigator-initiated study would not appear. The bound is therefore on the registry, not on all human exposure. The rectal-cancer trial is NOT YET RECRUITING with an estimated start date, so nothing is known about feasibility or tolerability of the dietary arm.

   ▸Evidence that argues against this (1)

   • The absence of a trial is an argument about pharmaceutical priorities and tool quality, not about the biology. NCT-503 and CBR-5884 are both explicitly described by their discoverers as tools: CBR-5884 is a noncompetitive inhibitor with time-dependent onset that disrupts PHGDH oligomerisation, discovered from an 800,000-compound screen, and NCT-503's own paper reports the unexpected result that PHGDH inhibition wastes one-carbon units from EXOGENOUS serine too - a mechanism that would not be predicted from the target and that changes what a clinical inhibitor would be expected to do. Pacold ME et al. 2016
8. E8established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   PubMed on 2026-09-22, one tag per term with translations read back: PHGDH[tiab] returns 665 records, of which 4 are fusion-scoped (29672864 Ewing/menin-SSP, 31235852 SOCS2 AML signature, 33301426 bone-marrow ATP/P2X7, 39579612 ailanthone KMT2A-MEN1 osteosarcoma) and 2 are menin-scoped; PHGDH[tiab] AND leukemia[tiab] returns 18; PGDH3[tiab] returns 6 and 3PGDH[tiab] returns 34, both with 0 fusion-scoped; and SERA[tiab] returns 135,340 with 26 fusion-scoped records that intersect PHGDH's four in ZERO ids.

   NCBI E-utilities esearch against PubMed, 2026-09-22, fusion operand set (KMT2A[tiab] OR MLL[tiab] OR "MLL-AF4"[tiab] OR "KMT2A::AFF1"[tiab] OR "MLL-rearranged"[tiab] OR "t(4;11)"[tiab] OR "MLL-AF9"[tiab] OR "MLL-ENL"[tiab]) with every cytogenetic term quoted and every term separately tagged · PubMed 2026
   ▸What this evidence cannot show

   A PubMed bound is bounded by TITLES AND ABSTRACTS, not by papers: supporting figures were not searched, and this graph already owns the worked counter-example (PMID 32766858 provides PHGDH-relevant evidence without naming PHGDH anywhere an index can see). The 4 fusion-scoped records were all opened by title; the 26 SERA records were not, because zero id overlap already establishes they are a different object. The 665 and 135,340 are positive counts and are therefore the class of figure rule 94 says to re-run if it becomes load-bearing - both were run in the one-tag-per-term form.

   ▸Evidence that argues against this (1)

   • The 2 menin-scoped hits include PMID 39579612, an ailanthone study targeting the KMT2A-MEN1 complex in osteosarcoma lung metastasis, which was not opened beyond its title and could contain a PHGDH measurement under menin-complex disruption in a third tumour type. A bounded rejection recorded rather than a read paper. Liang J et al. 2025
9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   In human and murine FLT3-ITD-driven AML, FLT3-ITD promotes serine synthesis and uptake via ATF4-dependent transcriptional regulation of de novo serine biosynthesis pathway genes and neutral amino-acid transport, and genetic or pharmacologic inhibition of PHGDH selectively inhibited proliferation of FLT3-ITD AMLs in vitro and in vivo and sensitised them to cytarabine.

   Serine Biosynthesis Is a Metabolic Vulnerability in FLT3-ITD-Driven Acute Myeloid Leukemia. · Bjelosevic S, Gruber E, Newbold A, et al., Johnstone RW · Cancer Discov 2021
   PMID 33436370DOI
   ▸What this evidence cannot show

   Abstract read this session; full text not retrieved, so the PHGDH inhibitor used, its concentrations and the effect sizes were not checked. **The system is FLT3-ITD AML, myeloid and not the index B-lineage KMT2A::AFF1 disease**, and the driver is a kinase mutation rather than a fusion. It is included because FLT3-ITD is a recurrent co-lesion in KMT2A-rearranged leukaemia and because it is the clearest published demonstration that an ATF4-driven serine programme in LEUKAEMIA is pharmacologically exploitable. It says nothing about the menin-selective component measured in [E1].

   ▸Evidence that argues against this (1)

   • This is an argument for PHGDH as a target in a co-mutation subgroup and simultaneously an argument against the fusion framing: if the serine programme is driven by whichever proliferative lesion the cell carries - EWS-FLI1, FLT3-ITD, a KMT2A fusion - then it is a generic feature of these states rather than a fusion-complex output, and the menin-selectivity measured here would need a different explanation. Jimenez JA et al. 2021

• depends on →MEN1 / menin — the adaptor the fusion cannot work without

  menin inhibition drops PHGDH -3.24 log2 in RS4;11 d7; H3K4me3 at the PHGDH promoter is menin-dependent in Ewing sarcoma

• regulates →ATF4 / CREB2 (P18848) - the hub this graph inferred from its targets and never wrote, and its own transcript does not move under either menin inhibitor

  ATF4 modulates core SSP transcription (PMID 33741715) - but PHGDH's compound-selectivity does not match the ATF4-output arm, so the input is not ATF4 alone

• related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)

  transcript collapse measured in three KMT2A::AFF1 lines (RS4;11, SEM, MV4;11); no dependency ever measured in this fusion

• ← inhibitsRevumenib (SNDX-5613, Revuforj) — first approved menin–KMT2A interface inhibitor

  class effect: VTP-50469 330 nM, RS4;11 d7, -3.24 log2; revumenib itself was not the compound tested

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

  EPZ5676 1 uM reproduces only 47% of the VTP-50469 effect on PHGDH, against 98-99% on the ISR-output genes

• part of →Serine/glycine dependence in MLL-ENL leukaemia, reached through a DIRECT fusion target gene - and the reason a PHGDH counted absence in this graph would have been wrong

  CBR-5884 is a PHGDH inhibitor; that entry's serine-dependency result is unreachable from the PHGDH symbol

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

  FLT3-ITD drives an ATF4-dependent serine programme and PHGDH inhibition is selectively active in FLT3-ITD AML

• supports →Menin inhibition collapses the serine/one-carbon/ASNS programme in KMT2A::AFF1 cells - PHGDH -3.24, PSAT1 -2.40, ASNS -1.72, flat in a wild-type line - so it should re-sensitise KMT2A-rearranged ALL to asparaginase, a combination PubMed has zero records for

  the serine arm of that hypothesis re-measured on a pre-registered panel; PHGDH -3.24 confirmed

• refutes →The ATF4/ISR collapse under menin inhibitors reports loss of the KMT2A-fusion transcriptional programme by ANY route, not engagement of the menin interface - and the loss is of ATF4 activity, not ATF4

  the serine arm is NOT compound-independent: PHGDH's VTP-minus-EPZ margin is -1.72 where the ISR arm's is -0.02 to +0.10, so the 'any route' claim does not extend from the ISR arm to the serine arm

• ← part ofPSAT1 (Q9Y617) - a DIRECT ATF4 target, and the gene that shows the brief's menin-selectivity claim was PHGDH's and not the pathway's

  steps 2 and 1 of the phosphorylated serine pathway - and they have different compound-selectivity (-0.55 vs -1.72), which is the beat's core observation

• ← supportsThe serine/one-carbon arm is NOT a pure ATF4 output: it is menin-selective where the ISR arm is compound-independent (median margin -0.55 vs -0.02, p=0.011), and it is ABSENT in KMT2A::MLLT3 AML where the ISR arm is present

  PHGDH carries the largest menin-selective margin in the panel, -1.72

• ← part ofALDH1L2 (Q3SY69) - the second-largest fall in the serine/one-carbon panel (-3.40), named nowhere in this graph, zero of 1,258 DepMap lines require it, and a tumour SUPPRESSOR where it has been studied

  mitochondrial one-carbon arm of the same programme; -3.40 against PHGDH -3.24 in RS4;11 day 7

• ← part ofSHMT2 (P34897) - the mitochondrial one-carbon step, and the paralogue that rises while it falls is why its loss may not matter

  mitochondrial one-carbon step downstream of serine synthesis; margin -0.42 against PHGDH -1.72

• ← refutesMTHFD1 (P11586) - the inverse case: the deepest leukaemia dependency in the serine/one-carbon panel, and the drug does not touch it

  within one panel, fold change and dependency are inverted: MTHFD1 +0.09 under the drug with 28.8% of lines below -0.5, PHGDH -3.24 with 1.1%

• ← refutesPSPH (P78330) - the third serine-synthesis enzyme, high in Ewing sarcoma alongside PHGDH and PSAT1, and the one that does NOT fall in leukaemia

  the three-enzyme pathway does not move as a unit: PHGDH -3.24, PSAT1 -2.40, PSPH -0.48, all detected, so no pathway-level claim survives

• ← depends onThe serine/one-carbon programme has a SECOND, chromatin-proximal menin input that is not ATF4 - which is why PHGDH is menin-selective where CHAC1 is not, and why the arm is absent in KMT2A::MLLT3

  PHGDH -1.72 margin is the strongest single measurement behind it, and the published H3K4me3 result is at the PHGDH promoter

• ← testsThe one experiment that settles this beat: dose-matched menin and DOT1L arms in RS4;11 with H3K4me3 ChIP at PHGDH, plus 13C-serine flux - and it answers the cycle-58 question at the same time

  the decisive number is the PHGDH log2 fold change in the dose-matched DOT1L arm