Oncofusion

start / graph / absence-enl-function-in-developing-brain

Explainerunrated confidence 0 of 5: unrated

ENL is the highest-expressed target in cerebellum and nobody has studied it there

also called MLLT1 cerebellum absence · ENL brain function unknown

Of all the drug targets in this project, ENL is the one most present in the part of the brain that grows fastest during a baby's first two years. Almost nothing has been published about what it does there.

The same thing, for a scientist

`ana-normal-tissue-exposure` put **MLLT1/ENL at 142 TPM in cerebellum**, its highest tissue and nearly three times its whole-body median of 52 — the largest brain value of any target in this graph. Two independent counted queries for its function there return **one record each**. There is an active `target-phase2-enl-yeats-dossier`, the disease is of the first year of life, and the cerebellum is the brain region that grows most in exactly that window. **This is a bounded absence, not a negative result.**

Go deeper
The queries and their counts

Run against PubMed on 2026-09-19:

queryrecords
MLLT1[tiab] AND (cerebellum[tiab] OR neurodevelopment*[tiab] OR brain development[tiab])1
ENL[tiab] AND cerebellum[tiab]1
(MLLT1[tiab] OR ENL[tiab]) AND knockout[tiab] AND mice[tiab]4
MLLT1[tiab] AND (germline[tiab] OR syndrome[tiab])7
YEATS domain inhibitor[tiab] AND toxicity[tiab]0

The single record returned by each of the first two is an exome study of consanguineous families with neurologic disease, not a study of ENL in the brain. 'ENL' is also a poor search token — it collides with unrelated abbreviations — so these counts are a floor on the real literature and the absence is bounded by the queries, not established.

What the counts do support is that no one has asked the question this project needs answered: what happens to a developing cerebellum when ENL is inhibited. Zero records for YEATS inhibitor toxicity says the same thing from the pharmacology side.

Why the absence matters more than the expression

A high TPM is a weak signal on its own and this graph says so repeatedly: mRNA is not protein and neither is dependency; menin sits high in cerebellum and menin inhibitors are tolerated. Taken alone, 142 TPM would not be worth an entry.

It is worth one because of what sits beside it. MLLT1 YEATS mutations cause a childhood kidney tumour by acting early in organ development (fusion-family-mllt1-yeats-wilms-tumour). So ENL is demonstrably consequential during development in at least one organ, is the most cerebellum-enriched target in this graph, and has an active inhibitor dossier aimed at infants — with no published work on either side of the question.

That combination is the definition of a gap this project is supposed to record rather than step over.

What germline constraint says, and why it settles less than it looks

One of this entry's open questions was whether gnomAD constraint would answer the developmental-tolerability question that GTEx cannot. It was run on 2026-09-19:

genepLILOEUFobserved / expected LoFtolerated pharmacologically?
KMT2A1.0000.07016 / 345.3n/a — haploinsufficiency causes Wiedemann-Steiner syndrome
MLLT31.0000.1724 / 53.2untested
MEN11.0000.2046 / 58.0yes — two approved inhibitors
PLK11.0000.2096 / 56.7inhibitors trialled, myelosuppressive
DOT1L1.0000.28833 / 153.3yes — pinometostat was tolerated
MLLT11.0000.33011 / 55.1untested
AFF10.7650.50248 / 121.7untested
CD190.0060.61532 / 69.8yes — blinatumomab, CAR-T

MLLT1 is strongly constrained — pLI 1.000, LOEUF 0.330, 11 observed loss-of-function variants against 55.1 expected. Read alone, that looks like confirmation of the worry.

It is not, and the table shows why. MEN1 is more constrained than MLLT1 on every measure, and germline heterozygous MEN1 loss causes a hereditary tumour syndrome — yet two menin inhibitors are approved and tolerated in humans. DOT1L is constrained and pinometostat was tolerated. The only unconstrained gene in the table, CD19, is also the one with the best-established infant tolerability.

So constraint and pharmacological tolerability agree once (CD19) and disagree twice (MEN1, DOT1L). A metric that has already failed on the drugs this project is built around cannot be used to raise doubts about the next one. Constraint measures lifelong germline haploinsufficiency; a drug is time-limited, dose-controlled, and given after the earliest developmental windows have closed.

The open question therefore survives its own answer. Neither expression nor constraint can tell us what inhibiting ENL does to an infant cerebellum — and that remains unstudied.

Facts and figures
absence classmeasured-not-analysed
absence class basisARCHIVE CHECK RUN 2026-09-21 (king). GEO holds 151 human developing-brain, fetal-brain or cerebellum single-cell RNA-seq DataSets, including GSE327101 (220 samples, 'Single-Cell Multimodal Atlas Reveals Regulatory Programs in the Human Brain'). Same split as the cell-state absences: the DESCRIPTIVE half - which neural populations and developmental stages express MLLT1/ENL, and whether that tracks the cerebellar enrichment this entry is built on - is a desk job on deposited data. The FUNCTIONAL half, what ENL DOES in developing brain, needs perturbation in that tissue and remains never-measured. Classified on the descriptive half because that is what the entry's own queries were searching for.
absence class assigned2026-09-21 (cycle 47, king)
archive checked
{
  "archive": "GEO DataSets",
  "date": "2026-09-21",
  "datasets_matching": 151,
  "verified_example": "GSE327101 (220 samples, human brain single-cell multimodal atlas)",
  "closes_the_descriptive_half": true,
  "closes_the_functional_half": false
}
How this could bite us back
  • This entry is a reason to look, not a reason to stop a programme. No evidence here shows ENL inhibition harms a developing brain; the point is that nobody has checked.
  • Do not cite the 142 TPM as a safety signal without the adult-tissue caveat. GTEx cannot measure an infant cerebellum, which is the only tissue the question is about.
  • The counted absence rests on two queries with a known-poor search token. Re-run it before quoting the gap to anyone making a decision.
Still open
  1. Re-run the counts with MeSH terms, the full MLLT1 alias set (ENL, YEATS1, LTG19, ENL/MLLT1) and a full-text rather than [tiab] search. A two-query floor is not a bounded absence this project should be satisfied with.
  2. Find a developmental expression atlas - BrainSpan or the Human Developmental Cell Atlas - and ask what MLLT1 does in the first two years rather than in adult post-mortem tissue.
  3. Constraint is answered and unhelpful: MLLT1 pLI 1.000, LOEUF 0.330, but MEN1 is more constrained still and its inhibitors are approved. The remaining instruments are a human loss-of-function phenotype and a developmental expression atlas.
  4. Query OMIM, ClinGen and DECIPHER for MLLT1 rather than PubMed. A germline phenotype would sit there under a condition name before it appears in a paper title, and the 11 observed loss-of-function carriers in gnomAD have never been phenotyped.
Every source behind this entry (4)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    Two independent PubMed queries for MLLT1/ENL function in the cerebellum or brain development return one record each, and that record is an exome study of consanguineous families with neurologic disease rather than a study of ENL in brain. A query for YEATS domain inhibitor toxicity returns zero records.

    What this evidence cannot show

    Counted 2026-09-19 with the exact queries recorded in this entry's first section. **'ENL' is a poor search token** — it collides with unrelated abbreviations and is not the primary symbol — and `[tiab]` searches miss a topic discussed only in a paper's body, so these counts are a floor. Per this project's standing rule an absence of testing is not a negative result: this records that the literature was not found by these queries, not that the work does not exist.

    Evidence that argues against this (2)
    • A two-query absence is exactly the shape this project has been burned by. A DOT1L recall figure of 3.7% turned out to be a five-fold query artefact, and the GEO PMID route returned a clean zero twenty-six times while being simply broken. These counts should be re-run with MeSH terms, the full alias set, and a body-text search before anyone treats the gap as real. NCBI 2026
    • Absence of a cerebellum-specific literature is unremarkable for most genes and proves nothing about safety in either direction. Thousands of genes are expressed in cerebellum at comparable levels with no dedicated literature, and singling this one out is driven by its dossier rather than by any signal in the data. GTEx Consortium / Broad Institute 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    GTEx v8 places MLLT1 at 142 TPM in cerebellum, its highest tissue, against a whole-body median of 52 across 54 tissues - the largest brain value of any target gene in this graph.

    What this evidence cannot show

    Queried 2026-09-19 via the GTEx API v2, datasetId gtex_v8. GTEx is adult post-mortem tissue; the infant cerebellum is not measured here and is the tissue the concern is actually about. 'Largest brain value of any target gene' is bounded to the 47 symbols this graph's target extractor resolved, which missed several real targets including AURKA.

    Evidence that argues against this (1)
    • Ranking highest among 47 targets is a claim about the sample, not about the cerebellum. MLLT1 is a broadly expressed transcription-elongation factor and its cerebellar enrichment is modest in absolute terms; several housekeeping genes exceed it severalfold in the same tissue. GTEx Consortium / Broad Institute 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    gnomAD loss-of-function constraint places MLLT1 among strongly constrained genes: pLI 1.000, LOEUF 0.330, with 11 observed loss-of-function variants against 55.1 expected. Human heterozygous loss of MLLT1 is selected against. The same query gives KMT2A LOEUF 0.070 (16/345.3), MLLT3 0.172, MEN1 0.204, PLK1 0.209, DOT1L 0.288, AFF1 0.502 (pLI 0.765) and CD19 0.615 (pLI 0.006).

    What this evidence cannot show

    Queried by the coordinator through the gnomAD GraphQL API on 2026-09-19 against GRCh38 and read off the response, not recalled. **Per this project's standing rule that a database fact is opposable on scope: pLI and LOEUF measure germline developmental constraint on heterozygous loss across a lifetime.** They are not a measure of whether time-limited pharmacological inhibition of the protein in a specific tissue is tolerated, which is the question the ENL dossier actually asks.

    Evidence that argues against this (2)
    • **Constraint does not predict drug tolerability, and this table contains the disproof.** MEN1 is maximally constrained (pLI 1.000, LOEUF 0.204, 6 observed LoF against 58 expected) and heterozygous germline loss causes multiple endocrine neoplasia type 1 — yet two menin inhibitors are approved and tolerated. DOT1L is constrained at LOEUF 0.288 and pinometostat was tolerated in trial. So a high constraint score for MLLT1 is not an argument against inhibiting it; if it were, it would have been an argument against the drugs this project is built around. Broad Institute gnomAD 2026
    • The one gene in this table whose therapy is unambiguously well tolerated in infants — CD19, pLI 0.006 — is also the only unconstrained one, so constraint and tolerability agree in exactly one case and disagree in two. Two agreements out of eight is not a relationship, and reading the MLLT1 row as a safety signal would be reading a metric that has already failed on this project's own drugs. GTEx Consortium / Broad Institute 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    No human germline MLLT1 loss-of-function phenotype is described in the literature reached by this query. All seven records answering `MLLT1[tiab] AND (germline[tiab] OR syndrome[tiab])` were read: two are bioinformatics transcription-factor network papers on polycystic ovary syndrome and type 2 diabetes where MLLT1 appears only as a network node; one reports variants of uncertain significance in a hypereosinophilic syndrome family; one is a veterinary study of a lethal calf syndrome; one is a CAR-T case report of a KMT2A::MLLT1 fusion; one is colorectal cancer epidemiology; and one is the COG/TARGET Wilms landscape. None describes germline MLLT1 haploinsufficiency in a person.

    What this evidence cannot show

    All seven abstracts fetched and read by the coordinator on 2026-09-19. **This is a bounded absence over one query, not an established negative** - the standing rule that an absence of testing is not a negative result applies with full force. A germline syndrome could be described under a phenotype name rather than the gene symbol, in OMIM or ClinGen rather than a titled paper, or in a case series whose abstract names neither 'germline' nor 'syndrome'. Neither OMIM nor ClinGen nor DECIPHER was queried.

    Evidence that argues against this (2)
    • The contrast with the neighbouring genes is what makes this informative and also what limits it. KMT2A haploinsufficiency causes Wiedemann-Steiner syndrome and germline MEN1 loss causes multiple endocrine neoplasia type 1 - both are well described, so the machinery for finding such syndromes clearly works in this gene family. But gnomAD gives MLLT1 11 observed loss-of-function variants against 55.1 expected: constrained, yet not zero. Eleven human heterozygous carriers exist and nobody has phenotyped them, which is a different situation from a gene whose loss is never seen. Broad Institute gnomAD 2026
    • Absence of a named syndrome is weak evidence of tolerability in either direction. Many strongly constrained genes have no syndrome because heterozygous loss is subtle or because nobody has looked, and a gene with no described phenotype is a gene with no described phenotype. NCBI 2026
Connected to