A noncoding element has no HGNC symbol, so every rule in this brief for resolving an identifier fails on it - worked on the one MYC enhancer that matters
also called the enhancer naming trap · BENC / BDME / unnamed · noncoding element identifier trap
Genes have official names, so when two papers call the same gene different things you can look it up and settle it. Stretches of regulatory DNA have no official names at all, so the same piece of the genome can appear in three papers under three different labels with nothing to connect them - and searching for any one label misses the other two.
The same thing, for a scientist
Brief rules 62 through 69 are all defences against gene-symbol ambiguity, and every one of them terminates in a database lookup: resolve the accession, read back `gene_primary`, assert `length_aa`, check the alias tables. **A noncoding regulatory element has no accession, no primary name and no alias table.** There is no HGNC for enhancers. The consequence is that the whole defence apparatus is structurally unavailable, and the failure is silent: a query on one of an element's names returns a clean, small, plausible-looking count with no signal that two other literatures exist. This entry works the trap on the single enhancer this beat needed - the MYC cluster 1.7 Mb downstream - where the measured numbers are as bad as they can be [E1][E2][E3].
The measurement
PubMed, run 2026-09-20, querytranslation read back on every query:
| query | records | signal |
|---|---|---|
"blood enhancer cluster"[tiab] | 3 | 3/3 on topic |
BENC[tiab] AND (MLL OR KMT2A) | 1 | 1/1 on topic |
BDME[tiab] | 14 | 0/14 on topic |
"BRD4-dependent MYC enhancer"[tiab] OR "BRD4-mediated MYC enhancer"[tiab] | 0 | the phrase is body-text only |
"MYC enhancer"[tiab] AND (MLL OR KMT2A) | 3 | 2/3 on topic |
The fourteen BDME records are Boswellia dalzielii methanol extract, bipolar disorder manic episode, breast duct microendoscopy, bis-demethoxycurcumin-PLGA microspheres, and zinc transport in bread wheat. This is brief rule 64's ordinary-word contamination in its purest form - the acronym is not a rare string, it is a common one that happens also to have been used once for an enhancer - and unlike "FACT" or "NELF" there is not even a partial signal to title-check out of the noise.
And the founding paper gives the element no name at all: Shi et al. describe it as 'a cluster of lineage-specific enhancers located 1.7 Mb downstream from Myc' 3. The only durable retrieval key is a coordinate, and a coordinate is not an indexable token.
Why this is a different failure from the gene-symbol traps already logged
The brief's existing traps have a common shape: one string, several referents (CD15 means FUT4 to HGNC and Lewis-x to this graph; MLL2 means two different genes; RING1 is a prefix of RING1B). The resolution in every case is a lookup that returns a canonical identifier.
This is the mirror: one referent, several strings, and no canonical identifier exists. There is nothing to look up. The three names were coined independently by three laboratories for three different reasons - BENC for what it does (blood enhancer cluster), BDME for what binds it (BRD4-dependent MYC enhancer), and nothing at all in the paper that found it - and no database reconciles them because databases reconcile symbols, not genomic intervals.
The practical damage is to counted absences, which this project relies on heavily. A beat that writes 'BENC has never been measured under a menin inhibitor, query X returns 0' has produced a technically true statement whose scope is one of three literatures. The zero is real; the bound is fiction. Brief rule 67 requires a positive control through the identical code path, and a positive control would have passed here - "blood enhancer cluster"[tiab] returns 3, the instrument works - while the absence claim remained unbounded. Rule 67 does not catch this class.
What to do instead
- For any noncoding element, enumerate names before counting anything. Read the full text of the paper that found it and the full text of at least one paper that cites it, and harvest every string they use for the element. Acronyms for enhancers are coined in body text and frequently never reach an abstract, so they are invisible to
[tiab]. - Search the coordinate description as free text.
"1.7 Mb" AND MYC AND enhanceris ugly and it is the only query that spans all three literatures here. - Search the target gene plus the word enhancer, which is the one formulation every author uses regardless of their private acronym:
"MYC enhancer"[tiab]returned the Bahr paper, the Rathert paper and one off-topic hit. - Title-check every acronym hit before counting it, and record the checked number.
BDMEwould otherwise contribute fourteen records to an apparent literature. - State the scope of the absence in terms of names searched, exactly as brief rule 62's septin case requires for retired gene symbols: 'zero records over the names BENC, BDME and the spelled-out phrase, searched 2026-09-20' - not 'zero records'.
The transferable version: when the object of a counted absence has no database identifier, the absence is bounded by the name list, and the name list must appear in the caveat.
| trap class | one referent, several strings, no canonical identifier |
|---|---|
| distinguished from | brief rules 62-69, all of which are 'one string, several referents' and all of which resolve by database lookup |
| why rule 67 does not catch it | rule 67's positive control tests whether the instrument can return a one. Here the instrument works perfectly - 'blood enhancer cluster'[tiab] returns 3 - and the absence is still unbounded, because the scope error is in the NAME LIST, not in the query mechanics. |
| worked example | noncoding-benc-myc-enhancer-cluster |
| tool | NCBI E-utilities esearch db=pubmed, querytranslation read back on every query |
| proposed brief rule | Rule 70 - a counted absence about a noncoding element is bounded by its NAME LIST, not by its query, and the name list goes in the caveat |
measured counts 2026 09 20 — the numbers behind the chart
{
"\"blood enhancer cluster\"[tiab]": 3,
"BENC[tiab] AND (MLL OR KMT2A)": 1,
"BDME[tiab]": 14,
"BDME[tiab] on topic": 0,
"\"BRD4-dependent MYC enhancer\"[tiab] OR \"BRD4-mediated MYC enhancer\"[tiab]": 0,
"\"MYC enhancer\"[tiab] AND (MLL OR KMT2A)": 3
}prescribed procedure
[ "enumerate names from the full text of the founding paper and of a citing paper before counting anything", "search the coordinate description as free text", "search target-gene + 'enhancer' as a phrase, which every author uses regardless of private acronym", "title-check every acronym hit and record the checked number, not the raw one", "state the absence's scope as the list of names searched, with the date" ]
- This concept undermines confidence in existing counted absences rather than adding any: any absence claim in this graph about an enhancer, written to the standard the brief currently prescribes, is narrower than it reads. Applying the rule retroactively will cost re-work.
- The prescribed procedure requires reading full texts before counting, which is materially slower than the current practice of querying first. A beat under time pressure will skip it, and the resulting absence will look identical to a correctly bounded one.
- How many other counted absences in this graph concern noncoding objects, and are any of them scoped to a single name? An audit of absence claims whose subject is an enhancer, lncRNA locus or CpG island would size the exposure.
- Is there any resource that reconciles enhancer names to intervals? ENCODE cCREs, SCREEN and the FANTOM enhancer atlas index intervals but not the names laboratories give them; if one of them does carry synonym fields, that would be the missing lookup and it should be found rather than assumed absent.
- Does the same problem apply to this project's own subnuclear and localisation vocabulary, where terms like 'Polycomb body' are similarly uncontrolled?
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The element is called the blood enhancer cluster (BENC) in the paper that named it, which characterises a conserved region 1.7 Mb downstream of Myc as a cluster of enhancer modules essential for Myc expression in haematopoietic stem cells and for maintenance of MLL-AF9-driven leukaemia in mice.
A Myc enhancer cluster regulates normal and leukaemic haematopoietic stem cell hierarchies. · Bahr C, von Paleske L, Uslu VV, Remeseiro S, Takayama N, Ng SW, et al. · Nature 2018What this evidence cannot show
The acronym BENC appears in this abstract, which is why it is retrievable at all; most enhancer acronyms are coined in body text and are not. This paper was not readable in full by this beat (no PMCID in Europe PMC), so its own name list for the element is known only from the abstract, which means the enumeration performed here is itself incomplete by the standard this entry sets.
Evidence that argues against this (1)
- The same element is called BDME in a paper that cites the founding work directly, so 'BENC' is not the community's settled name and a BENC-only search is not a search of the element's literature. That is the finding, and it is its own counter-evidence to any BENC-scoped absence claim. Pulikkan JA et al. 2018
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The same element is called BDME - 'BRD4-dependent MYC enhancer' in the introduction and 'BRD4-mediated MYC enhancer' in the results of the same paper - located 1.7 Mb downstream of the MYC TSS with five conserved elements E1 to E5, cited to Shi et al. 2013; the acronym appears only in the body text and not in the abstract.
CBFbeta-SMMHC Inhibition Triggers Apoptosis by Disrupting MYC Chromatin Dynamics in Acute Myeloid Leukemia · Pulikkan JA, Hegde M, Ahmad HM, Belaghzal H, Illendula A, Yu J, O'Hagan K, Ou J, Muller-Tidow C, Wolfe SA, Zhu LJ, Dekker J, Bushweller JH, Castilla LH · Cell 2018What this evidence cannot show
Read from the PMC author manuscript (PMC6211564) on 2026-09-20 via the PMC HTML route, tags stripped before entities were unescaped. The paper itself uses two spellings of the same acronym's expansion within one article, which is a smaller instance of the same problem. It does not state that BDME and BENC are the same object - it predates neither but cites Shi rather than Bahr - so the equivalence is this graph's reading of two coordinate descriptions, not a quotation.
Evidence that argues against this (1)
- Because neither paper asserts the equivalence in words, a reader could hold that BDME and BENC are overlapping but non-identical intervals. Nothing in either abstract excludes that. The equivalence rests on both being described as 1.7 Mb downstream of the MYC TSS and on Pulikkan citing Shi, which is the same element Bahr renamed; it is a strong reading and it is a reading. Bahr C et al. 2018
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The paper that discovered the element gives it no name, describing it only as 'a cluster of lineage-specific enhancers located 1.7 Mb downstream from Myc' occupied by SWI/SNF and BRD4.
Role of SWI/SNF in acute leukemia maintenance and enhancer-mediated Myc regulation. · Shi J, Whyte WA, Zepeda-Mendoza CJ, Milazzo JP, Shen C, Roe JS, et al. · Genes Dev 2013What this evidence cannot show
Established at abstract level: the abstract contains the coordinate description and no acronym. The body text was not read for this entry, so it is possible - though it would not change the retrieval problem - that a name appears there. Any name in body text is by definition invisible to a [tiab] search, which is the point.
Evidence that argues against this (1)
- An unnamed element is not a retrieval failure the authors could have avoided by naming it, because the two names that were subsequently coined do not retrieve each other either. Naming is not the fix; enumerating names before counting is. This citation is therefore evidence of the structural problem rather than of an authorial omission. Watanabe A et al. 2024
the worked example: BENC, BDME, and no name at all
- related to →Mechanism class 5 — enhancer hijacking (inv(3) GATA2/EVI1 — and DUX4-r, which only half belongs here)
any counted absence about an enhancer in this graph is bounded by its name list, not by its query
the beat's methodological finding: rule 67's positive control does not protect an absence about a noncoding element