Metabolic enzyme enumeration in KMT2A-rearranged leukaemia: the instrument, its positive controls, seventeen bounded absences - and the counted zero that would have been wrong

Tool: Europe PMC REST search endpoint, resultType=core, pageSize=1000, run 2026-09-20.

Query form: free text, <SYMBOL> AND (MLL OR KMT2A) AND leukemia. Free text was chosen deliberately because rule 67 records that the field-scoped form (TITLE:"X" OR ABSTRACT:"X") AND (TITLE:"KMT2A" OR ...) does not compose against this API and returns 0 regardless of the literature.

Title check, version 1 (BROKEN): require the gene symbol in the title and KMT2A/MLL in the title. Run against the positive controls this returned 0 for IMPDH2 - a gene with a 2025 Nature Communications paper directly on this beat. The reason is that the paper is titled Guanine nucleotide biosynthesis blockade impairs MLL complex formation... and never names IMPDH2 in its title. The instrument was rejected before any absence was recorded.

Title check, version 2 (USED): require the gene symbol in title or abstract AND KMT2A/MLL/11q23/'mixed lineage leukemi' in the title. Positive controls on this version, same day, same code path:

Three controls returning non-zero is the licence to record a zero.

The lesson is not that version 1 was careless. It is that a title-only check is a perfectly reasonable design that silently fails for the exact papers a metabolism beat most wants - those that frame a metabolic intervention at the level of the pathway or the complex rather than the gene. Rule 64 records this for complex subunits. This is the same failure for pathways named by their product or their inhibitor.

All measured 2026-09-20, Europe PMC REST, query <SYMBOL> AND (MLL OR KMT2A) AND leukemia, instrument version 2, full result set read (not a first page).

Reading these correctly. A zero here means: no paper carries this symbol in its title or abstract while naming KMT2A or MLL in its title, in Europe PMC, on 2026-09-20, on this instrument. It does not mean nobody has studied the enzyme in this disease, for the reason set out next.

PHGDH returns 0 on the instrument above. Writing that as 'serine biosynthesis is unstudied in KMT2A-rearranged leukaemia' would have been the beat's worst error, because PMID 32766858 tests exactly that and reports that inhibition of Ser biosynthesis by CBR5884 had a significantly stronger influence on self-renewal and survival of transformed cells than on normal controls, with ROS, DNA damage and apoptosis predominantly in leukaemic cells, and a Ser/Gly-free diet significantly extending survival in transplantation experiments 1.

CBR5884 is a PHGDH inhibitor. The paper is unreachable from the symbol because its abstract frames the intervention as Ser biosynthesis and names the compound, never the enzyme.

The transferable rule, offered for the brief: before recording a counted absence on a metabolic enzyme, run the inhibitor name and the pathway phrase as well as the symbol. For this beat's candidate list that means at least:

This is the same family as rule 64 (complex subunits are unreachable from subunit symbols) and the retired-symbol rule (SEPT9 versus SEPTIN9). The new member is: a pathway is often named by its drug.

Recorded because the brief's standing instruction is that a paper a reader would expect to be cited, and which is not, should leave a trace.

• PMID 39789150, Lactylation modulation identifies key biomarkers and therapeutic targets in KMT2A-rearranged AML (Sci Rep 2025). Squarely on-beat by title - lactate is the glycolytic end product and lactylation is a metabolism-to-chromatin link this beat would want. Rejected because the study is a machine-learning analysis of public expression data that outputs a six-gene prognostic panel (PFN1, S100A6, CBR1, LDHB, LGALS1, PRDX1) with no functional validation and no lactate measurement; the named genes are not lactylation enzymes. What would change my mind: any functional test of lactylation writers or erasers in KMT2A-rearranged cells.
• PMID 42180876, NAMPT/NAD salvage in AML. Rejected as a source: the PubMed record is typed Editorial/Comment and returns an empty abstract; the primary study behind it was not identified. What would change my mind: the primary paper's identifier.
• PMID 41730833 (marrow leptin-LEPR rewiring mitochondrial oxidative metabolism) and PMID 40501394 (PRDM16 metabolic reprogramming and cytarabine resistance). Both real metabolic AML papers; rejected as not KMT2A-stratified on their abstracts.
• PMID 41657788, matrine plus RSL3 ferroptosis in MLL-rearranged AML. Not rejected - unread. Title only; the indexed title contains a probable transcription error ('RLS3' for RSL3). It is a reading-queue item and is recorded at weak confidence in gene-acsl4.
• GLS / glutaminase. Enumerated as a candidate because glutamine is named in 5 graph entries and the brief flagged it. Dropped: 127 joined records, 0 passing the title check, and no KMT2A-stratified glutaminase experiment found. What would change my mind: a CB-839 experiment in KMT2A-rearranged material.

ACTB (46 entries): do not write an entry. Confirmed by reading the contexts. Word-boundary matches in this graph fall into two classes, both non-biological: (i) the negative control row in this project's own DepMap analyses - ACTB (control) -0.9493 ... the per-line offset floor, and ACTB r = 0.15 (CI crosses zero) beside GAPDH in a co-dependency null; (ii) the qPCR housekeeping amplicon in the Swedish TREC/KREC/ACTB newborn SCID screening programme. In no entry is ACTB a biological subject. The brief's guess was right and the correct outcome is a recorded decision not to write it.

PTH (37 entries): not a metabolism-beat item, and resolving it to a gene would be a category error. Every context read is the serum analyte - intact PTH raised in 34% of patients on the revumenib label, PTH as an absent endocrine outcome measure in three trial protocols, raised PTH with hypercalcaemia in MEN1 syndrome. The graph is talking about a pharmacodynamic and toxicity measurement, not about the PTH gene in a leukaemic blast. This belongs to the endocrine beat, which already covers it.

That is a generalisable warning about the named-but-unwritten queue itself. The queue counts tokens, and a token can be a gene, a reagent control (ACTB), or a clinical assay analyte (PTH). Writing gene-actb or gene-pth on queue position alone would produce two confident, well-cited entries about objects this graph is not discussing - the same class of error as rule 69's CD-number case, where a token resolves to a gene that is not what the graph means.