AFF1-AS1 - the antisense RNA 151 bp from the index fusion partner's TSS, which Ensembl does not recognise and this project's promoter table cannot see

A database record fetched 2026-09-24, not a paper. The absence of ensembl_gene_id is an assertion about THIS record as returned on this date; HGNC adds Ensembl mappings over time and this one may acquire it. The GeneID was resolved by esearch on AFF1-AS1[sym] AND human[orgn] (querytranslation read back as 'AFF1-AS1[sym] AND "Homo sapiens"[Organism]', count 1), not typed from memory or copied from a sibling entry.

• An HGNC approval is a nomenclature decision, not a demonstration that a transcript is made, is stable, or does anything. HGNC approves antisense symbols on annotation evidence alone. So 'AFF1-AS1 is a gene' is a statement about a naming committee, and the biology behind it is exactly what E4 and E5 show is absent. NCBI Gene 2026

UCSC txStart/chromStart are 0-based half-open and txEnd is the 1-based end, so the minus-strand TSS is txEnd and the plus-strand TSS is txStart; the 151 bp is the 0-based difference and the TSS-to-TSS distance counted in 1-based coordinates is 152. The 92.45% is a property of this project's 2 kb window convention as much as of the genome: a 500 bp window would share 69.8% and a 5 kb window 97.0%. refGene here is the same snapshot the project's promoter table is built from (dataTime 2020-08-17 per Q153), which is why it is the right annotation for describing what the project actually measures and the wrong one for describing the genome today.

• Divergent promoter pairs are ordinary: bidirectional promoters separated by under 1 kb were estimated at more than 10% of human genes, so this geometry is not special to AFF1 and does not by itself imply shared regulation. What it implies is a measurement limit, which is the only claim made here. Trinklein ND et al. 2004

Ensembl REST returned HTTP 500 and HTTP 503 on four attempts during this beat before serving these records, and a failed fetch is an instrument failure rather than an absence, so every statement here rests on a response that was actually received and saved (q155_ens_region.json, q155_ens_tx.json). The empty xrefs result is NOT treated as an instrument failure because the identical call on two control symbols returned records in the same session. Whether ENSG00000284968 and AFF1-AS1 are 'the same gene' is a curation question nobody has adjudicated; they share the locus and the antisense orientation and differ by 1,352 bp at the 5' end.

• The directional claim - that Ensembl puts the antisense TSS inside AFF1 rather than divergent from it - is true of three of Ensembl's four transcripts and FALSE of the fourth, ENST00000509262, whose TSS at 86,934,542 sits 459 bp inside the project's AFF1 window on the divergent side. So the two annotations are not simply in opposition; Ensembl contains RefSeq's reading as a minority model. Quoting only the 0.05% gene-level overlap would overstate the disagreement. UCSC Genome Browser et al. 2026

This is a bound on a NAME LIST, not on the biology. The names tried were AFF1-AS1, AFF1-AS, 'AFF1 antisense', 'AFF1 antisense RNA 1', LOC100506746, 'AF4 antisense', 'MLLT2 antisense'; a paper describing an uncharacterised antisense transcript by coordinate, or calling it something else entirely, would be invisible to all of them. Two of the queries this script labelled 'controls' - "IRX2-DT"[tiab] and IRX2[tiab] AND antisense[tiab] - returned zero; they were mislabelled by me and are comparators, not controls, and the genuine positive controls (AFF1 226, AF4 1219, MLLT2 6, HOTAIR+antisense 633, "HOXA-AS2" 99, KMT2A+lncRNA 9) all returned large values through the same shapes, which is why the zeros above are recorded as real. The GEO query is the archive check rule 47 requires and it returned quotedphrasesnotfound rather than a bare zero, which is the stronger form of negative.

• A zero in PubMed and Europe PMC bounds what has been REPORTED, not what has been measured. Any RNA-seq of a KMT2A::AFF1 sample contains reads over this locus; the absence is measured-not-reported at worst and measured-not-analysed at best, not never-measured. The cheap closing move is a stranded RNA-seq lookup in an existing deposit, not a new experiment. NCBI GEO DataSets / E-utilities 2026

A review is a map, not a primary source, and this one is cited here PRECISELY for what it does not contain, which is the weakest kind of citation unless the instrument is stated. The instrument: Europe PMC fullTextXML for PMC12922646, HTTP 200, 176,836 bytes, <body> present, 56,262 characters after stripping tags first and unescaping entities second, and EVERY unicode dash (U+2010, U+2011, U+2012, U+2013, U+2014, U+2212) normalised to ASCII hyphen before searching - the document uses U+2010 and an ASCII-hyphen search for 'AFF1-AS1' would have returned a meaningless zero for a different reason. A review can omit something that is published; this bounds the field's attention, not the literature.

• The review does not name C4orf36 either, and C4orf36 IS a protein-coding gene with a current-RefSeq TSS 269 bp from AFF1's. So the review's silence is about the whole neighbourhood rather than specifically about an antisense RNA, which weakens any inference that AFF1-AS1 in particular has been considered and dismissed. UCSC Genome Browser et al. 2026

There is no spike-in in this deposit, so every value is a within-sample share and cannot separate low target abundance from a shallow library. The run is complete - 133 of 133 libraries, zero failures, asserted before any verdict was computed - but the floor is the maximum over eight tissue-restricted loci at the same width, which is a conservative floor and not a null distribution. A ratio of body means over a 9,957 bp interval says nothing about where inside the interval the signal is; the base-resolution profile in the sibling entry does, and it does not support the directional reading this ratio suggests.

• A zero in AFF1-AS1's gene body is weak evidence that the promoter signal is AFF1's, because the assay is unstranded protein occupancy on DNA, not transcription. A divergent promoter is ONE piece of DNA, and an occupancy measurement at it has no gene to belong to. The gene-body asymmetry tests whether an elongation complex travels into one transcription unit, which is a different and narrower question. Core LJ et al. 2008