"ENL" returns 1,109 PubMed records of which about 30% are the protein: a measured alias-precision failure
also called ENL token collision · alias precision MLLT1 · erythema nodosum leprosum collision
The short name ENL is used for at least three unrelated things in the medical literature - the leukaemia protein, a skin reaction in leprosy, and a plant compound. Searching for it returns mostly the wrong ones, so any count built on that word is wrong by roughly threefold unless you check every hit.
The same thing, for a scientist
This project's standing alias rule has two halves: searching a retired symbol prevents under-counting, and searching a short or word-like alias causes over-counting. MLLT1 is a clean worked example of the second half, measured rather than asserted. `ENL[tiab]` returns **1,109** PubMed records; `MLLT1[tiab]` returns 94, `YEATS1[tiab]` 3 and `LTG19[tiab]` 9. A 60-record sample of the `ENL[tiab]` result set, classified by title and abstract, breaks down as **18 MLLT1/leukaemia-context (30.0%), 21 erythema nodosum leprosum (35.0%), and 21 unrelated (35.0%)**. The unrelated class is dominated by **enterolactone**, a plant lignan metabolite abbreviated ENL in nutrition and metabolism papers, and includes radar image-enhancement and SAR remote-sensing records where ENL is 'equivalent number of looks'. Extrapolated, roughly 333 of 1,109 records concern the protein. **Any counted-absence or counted-presence claim built on `ENL[tiab]` without title checking is wrong by about 3.3-fold**, and it will read as a healthy, confident number.
The measurement
| token | records |
|---|---|
ENL[tiab] | 1,109 |
MLLT1[tiab] | 94 |
LTG19[tiab] | 9 |
YEATS1[tiab] | 3 |
Sample of 60 records from ENL[tiab], classified by reading title and abstract:
| class | n | % |
|---|---|---|
| MLLT1 / super elongation complex / leukaemia | 18 | 30.0 |
| erythema nodosum leprosum | 21 | 35.0 |
| unrelated (enterolactone, remote sensing, other) | 21 | 35.0 |
The unrelated class is not noise in the usual sense; it is a second and third homonym with substantial literatures of their own. Enterolactone appears in cardiovascular, oncology and gut-microbiome papers, several of which also contain cancer vocabulary, so a query combining ENL with cancer terms does not clean the set. That is what makes this collision worse than an ordinary one.
Which direction it breaks a claim
Over-retrieval and under-retrieval fail in opposite directions and both produce a confident wrong number.
- A counted presence built on
ENL[tiab]inflates. 'There are over a thousand papers on ENL' would be roughly threefold wrong. - A counted absence built on
ENL[tiab]is safe against this particular failure, because adding noise can only raise a count: if a query returns zero after including a noisy token, it returned zero despite the noise.absence-enl-yeats-and-condensate-phospho-gatingrelies on exactly that asymmetry and says so.
The practical rule this beat used: include ENL in absence queries, exclude it from presence counts, and title-check anything in between.
The other partners, for comparison
| partner | current symbol | records | word-like alias | records | verdict |
|---|---|---|---|---|---|
| MLLT1 | MLLT1[tiab] | 94 | ENL[tiab] | 1,109 | catastrophic collision (3 homonyms) |
| MLLT3 | MLLT3[tiab] | 244 | AF9[tiab] | 807 | heavy collision; AF9 is also used as a fusion NAME (MLL-AF9), so most hits are AML papers that never study the AF9 protein |
| MLLT10 | MLLT10[tiab] | 178 | AF10[tiab] | 280 | moderate collision (B-cell line AF-10, C. elegans ZFP-1/AF10) |
| ELL | - | - | ELL1[tiab] | 22 | ELL[tiab] alone is unusable; paired with elongation/transcription terms it is workable |
| ELL2 | ELL2[tiab] | 91 | - | - | clean |
The AF9 case is a distinct trap worth naming separately. (MLLT3 OR AF9 OR "AF-9")[tiab] AND (phosphorylation OR phosphorylated OR kinase OR phosphosite)[tiab] returns 106 records, and title inspection of the first 40 shows they are almost entirely MLL-AF9 leukaemia papers in which 'phosphorylation' refers to oxidative phosphorylation, STAT phosphorylation, or a signalling pathway - not to phosphorylation of the AF9 protein. The fusion's name contains the partner's name, so the partner's protein literature is buried inside the disease literature. A beat searching for AF9 phosphorylation and reading the count as 106 would conclude the topic is well studied; the true number of records mapping a site on AF9 is one.
| status | run |
|---|---|
| rule used in this beat | include ENL in absence queries (noise can only raise a count, so a zero survives it), exclude it from presence counts, title-check everything between |
| af9 trap | the fusion name MLL-AF9 buries the partner's protein literature inside the disease literature; 106 records for AF9-plus-phosphorylation, 0 of the first 40 mapping a site |
| result summary | ENL[tiab] over-retrieves about 3.3-fold against three homonyms. AF9 fails differently: the symbol matches the fusion name, so the partner's protein literature is unreachable from the obvious query. |
token counts — the numbers behind the chart
{
"ENL[tiab]": 1109,
"MLLT1[tiab]": 94,
"LTG19[tiab]": 9,
"YEATS1[tiab]": 3,
"MLLT3[tiab]": 244,
"AF9[tiab]": 807,
"\"AF-9\"[tiab]": 110,
"MLLT10[tiab]": 178,
"AF10[tiab]": 280,
"\"AF-10\"[tiab]": 184,
"ELL1[tiab]": 22,
"ELL2[tiab]": 91
}enl sample — the numbers behind the chart
{
"n": 60,
"mllt1_or_leukaemia": 18,
"erythema_nodosum_leprosum": 21,
"unrelated": 21,
"protein_share_pct": 30,
"estimated_true_records": 333,
"inflation_factor": 3.3
}homonyms of ENL
[ "MLLT1 protein", "erythema nodosum leprosum", "enterolactone", "equivalent number of looks (SAR remote sensing)" ]
- This entry's practical consequence is negative and specific: any counted-presence number in this graph built on ENL or AF9 as a search token should be treated as unreliable until title-checked, and a well-written absence claim is exactly the kind that looks most trustworthy while being wrong.
- Does the same fusion-name burial affect MLLT1? MLL-ENL is a common fusion name, and the ENL homonym problem may be masking a second, independent retrieval failure of the AF9 kind.
- How many of this graph's existing counted absences were run over word-like aliases without title checking? That is an auditable question across the repository.
- Would a MeSH- or gene2pubmed-keyed retrieval outperform symbol queries for these partners? Neither was tried here.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
PubMed returns 1,109 records for ENL[tiab], 94 for MLLT1[tiab], 9 for LTG19[tiab] and 3 for YEATS1[tiab]; in a 60-record sample of the ENL[tiab] set classified by title and abstract, 18 concern MLLT1 or leukaemia, 21 concern erythema nodosum leprosum and 21 are unrelated, principally enterolactone and remote-sensing usages.
PubMed esearch and efetch, run 2026-09-19; 60-record sample of ENL[tiab] classified by regular-expression screening of title plus abstract followed by inspection of the unmatched records · NCBI E-utilities 2026What this evidence cannot show
The sample is PubMed's default ordering (most recent first), not a random sample, so the proportions reflect the recent literature and may not hold across the whole 1,109. Classification was by keyword screen with manual inspection of the residual, so borderline records could fall in the wrong class. The 30% figure is an estimate with a sample-size uncertainty of roughly plus or minus 12 percentage points at 95% confidence and should be quoted as 'about a third', not as 30.0%.
Evidence that argues against this (1)
- Recency biases the estimate: enterolactone is a currently active nutrition field while the MLLT1 literature is steady, so a sample from the most recent records probably understates the protein's share across the full set. The direction of the bias is known and the conclusion - that ENL[tiab] cannot be used as a presence count - survives it either way. NCBI E-utilities 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
A query for MLLT3/AF9 phosphorylation returns 106 records of which title inspection of the first 40 finds none that maps a phosphorylation site on the AF9 protein, because the fusion name MLL-AF9 makes the partner's symbol match the disease literature and 'phosphorylation' in those records refers to oxidative phosphorylation or to signalling pathways.
PubMed esearch and efetch 2026-09-19: (MLLT3[tiab] OR AF9[tiab] OR "AF-9"[tiab]) AND (phosphorylation OR phosphorylated OR kinase OR phosphosite)[tiab] = 106 records; first 40 titles inspected · NCBI E-utilities 2026What this evidence cannot show
Only the first 40 of 106 were title-checked, so a mapping paper could sit in the unchecked 66. The one paper that does map an AF9 site was found by a different route - a YEATS-plus-phosphorylation query and a super-elongation-complex query - which is itself the argument that a single query formulation is never enough.
Evidence that argues against this (1)
- The AF9 site-mapping paper exists and was found, which shows the collision is a retrieval problem rather than an absence. A beat that ran only the symbol query and stopped would have reported a false absence for the partner where this beat's headline result lives. Talukdar P et al. 2024
why the zero counts survive the noisy token
ENL token collides with three unrelated literatures
- related to →MLLT3 / AF9 — a famous fusion partner that is not a dependency, and a hub domain that is
AF9 symbol is buried inside the MLL-AF9 disease literature
same title-checking discipline applied to AF10
alias precision measured for this beat