C4orf36 - a protein-coding gene with three current TSSs spanning 43,782 bp, two of them inside or beside AFF1's promoter window, and the project's table has room for one
also called MGC26744 · chromosome 4 open reading frame 36 · ENSG00000163633 · ENSG00000285458
A third gene starts inside the stretch of DNA this project uses to measure AFF1's on-switch, and it is protein-coding. It does have a row in the project's table, and that row is not wrong - the gene has three different start sites 43,000 letters apart and a one-row-per-gene table can record only the one furthest from AFF1.
The same thing, for a scientist
C4orf36 is **NCBI GeneID 132989**, **HGNC:28386**, 4q21.3, protein-coding, UniProt **Q96KX1** (CD036_HUMAN, *Uncharacterized protein C4orf36*, 117 aa) [E1]. In the **current** UCSC hg38 `ncbiRefSeqCurated` track it has **twenty** curated transcripts at **three distinct TSSs** on the minus strand [E2]: | TSS | transcripts | distance to AFF1's TSS | |---:|---:|---:| | **86,892,422** | 13 (`NM_144645.4`, `NM_001369888/889`, `NM_001414635/636/638`, `NR_183013-183019`) | 42,579 bp | | **86,934,732** | 4 (`NM_001414641.1`, `NM_001414642.1`, `NR_183021.1`, `NR_183022.1`) | **269 bp - inside the project's AFF1 window** | | **86,936,204** | 3 (`NM_001414639.1`, `NM_001414640.1`, `NR_183020.1`) | 1,203 bp - 203 bp outside the window | The project's promoter table scores C4orf36 at `chr4:86,891,422-86,893,422`, keyed on `NM_144645`. 🔴 **CORRECTED: that row is NOT displaced and NOT stale.** This entry originally said the row was *"42,310 bp from where current RefSeq puts it"*. **It is wrong.** `86,892,422` is one of the three current TSSs and is the one supported by 13 of the 20 curated transcripts; a whole-table audit run immediately after this beat measured its `d_nearest` as **0** [E3]. The error was mine and its mechanism is worth more than the number: I computed the displacement from a UCSC track query restricted to `chr4:86,915,000-86,945,000`, **a window that by construction cannot contain the 86,892,422 transcript**, and the query returned a confident 42,310 bp displacement because the answer had been filtered out before the comparison ran. **What survives, and is the real finding:** C4orf36 has **three equally current TSSs spanning 43,782 bp**, and this project's promoter table has **one row per symbol**. It keeps the distal one. **Two other current C4orf36 TSSs sit 269 bp and 1,203 bp from AFF1's**, and neither is visible to any within-table check. The table is not wrong; it is **one of three right answers, and there is room for only one**. The whole-table audit puts this class at **14.69%** of rows - 2,851 symbols with multiple non-overlapping TSSs, median span 21,291 bp - against **1.83%** for genuine annotation drift [E3]. **Ensembl carries both ends under one symbol.** `ENSG00000163633` (HGNC-sourced) is the short gene at `chr4:86,876,205-86,892,813`; `ENSG00000285458` (NCBI-sourced) is the long one at `chr4:86,886,472-86,936,178`, reaching AFF1's promoter. **HGNC's own `ensembl_gene_id` points at the short one** [E4]. So a symbol-keyed coordinate lookup returns either, with no error. **The measured consequence:** the project's C4orf36 row is occupied in **0 of 27** GSE252378 samples, while a 400 bp window on the `86,936,204` TSS is occupied in **23 of 27** and carries the **highest KMT2A-N-over-KMT2A-C ratio anywhere in the 6 kb around AFF1 (N/C 3.45)** [E5].
Three TSSs, one row - and the correction that got this right
Fetched from UCSC ncbiRefSeqCurated over chr4:86,800,000-87,000,000 after the whole-table audit contradicted this entry's first version:
13 transcripts txEnd 86,892,422 (NM_144645.4, NM_001369888/889, NM_001414635/636/638, NR_183013-19)
4 transcripts txEnd 86,934,732 (NM_001414641/642, NR_183021/022)
3 transcripts txEnd 86,936,204 (NM_001414639/640, NR_183020)
All minus strand, so the TSS is txEnd. All curated (NM_/NR_), none predicted.
The original query that produced the wrong answer was the same track over chr4:86,915,000-86,945,000. It returned seven C4orf36 transcripts at two TSSs and none at 86,892,422 — correctly, because that coordinate is 23 kb below the window's lower edge. The script then compared the project's row TSS against min(observed TSSs) and reported a 42,310 bp displacement. Every step was correct and the conclusion was false, because the filter that defined the query removed the case that would have refuted it. A filter must report what it removed; this one silently defined the answer.
The right framing, which the whole-table audit supplies: d_nearest = 0, displaced = false, n_tss = 3, span = 43,782 bp.
What is known about the protein: almost nothing, and nothing in leukaemia
UniProt Q96KX1, CD036_HUMAN, recommended name "Uncharacterized protein C4orf36", 117 aa, reviewed. Three unreviewed TrEMBL fragments (D6RAE1 81 aa, D6RGW9 113 aa, D6RCP1 46 aa) also carry C4orf36 as primary gene name; the reviewed record is the one asserted here and the length is stored so verify-accessions.py can catch a swap.
PubMed, this cycle: C4orf36[All Fields] returns 3 records - an alcohol-associated-liver-disease TWAS (PMID 40823550), a Circ_C4orf36 BMSC osteogenesis paper (PMID 36242722) and a trophoblast RNA profiling study (PMID 34557907). C4orf36[All Fields] AND leukemia[All Fields] returns 0 (querytranslation expands leukemia through MeSH, so the zero is against a wide net and is conservative). C4orf36[All Fields] AND AFF1[All Fields] returns 1, the TWAS, where both are locus labels.
Nobody has written about this gene in a haematological context, and nobody has written about its position next to AFF1 at all.
Why this matters to this project specifically
Q153 established that 21.6% of this project's 19,404 promoter windows contain another gene's TSS, and identified AFF1's as one of the worst cases because AFF1-AS1 is structurally invisible to an NM_-derived table. C4orf36 is a second, different way the same check fails:
- AFF1-AS1 has no row, because it has no
NM_transcript, so a within-table neighbour check returns a clean no neighbour. - C4orf36 has a row, at a legitimate current TSS, and the check returns a clean no neighbour for a different reason - the recorded TSS is one of three and happens to be the one 42.6 kb away.
Both produce the same confident negative, and neither is detectable by auditing the table against itself. The remedies are different too: the first needs the full annotation rather than the NM_ subset; the second needs all of a symbol's TSSs, not one, and cannot be fixed by refreshing the table at all.
The measured consequence is in ana-q155-the-fusion-biased-peak-at-the-aff1-locus-fires-in-every-partner-so-it-is-not-autoregulation: at chr4:86,936,001-86,936,401, centred on the 86,936,204 TSS, the KMT2A-N mean is 15.24 against a KMT2A-C mean of 4.42 (N/C 3.45) and 23 of 27 samples clear the occupancy rule, while the project's C4orf36 row clears it in 0 of 27. The table does not merely fail to see the near TSS; it reports the symbol as empty.
| ncbi gene ID | 132989 |
|---|---|
| hgnc ID | HGNC:28386 |
| gene primary | C4orf36 |
| description | chromosome 4 open reading frame 36 |
| locus type | gene with protein product |
| cytoband | 4q21.3 |
| uniprot | Q96KX1 |
| uniprot entry name | CD036_HUMAN |
| protein name | Uncharacterized protein C4orf36 |
| length AA | 117 |
| resolved how | UniProt REST gene_exact:C4orf36 AND organism_id:9606, fetched 2026-09-24. Four records returned; the single REVIEWED one (Q96KX1) was kept after reading gene_primary back, and the length is asserted so verify-accessions.py can catch a swap. The three unreviewed fragments D6RAE1 (81 aa), D6RGW9 (113 aa) and D6RCP1 (46 aa) are recorded and not used. |
| n curated transcripts | 20 |
| tss span bp | 43782 |
| in project promoter table | true |
| project promoter window | chr4:86,891,422-86,893,422 (TSS 86,892,422 +/- 1000, keyed on NM_144645) |
| project row is displaced | false |
| project row d nearest bp | 0 |
| correction cycle 142 | WITHDRAWN the same cycle it was written. This entry's first version said the project's row was 'displaced 42,310 bp from where current RefSeq puts it'. FALSE: 86,892,422 is one of three current TSSs and is the one carrying 13 of the 20 curated transcripts, so d_nearest = 0. The error came from computing the displacement against a UCSC track query restricted to chr4:86,915,000-86,945,000, a window that by construction excluded that transcript. Found by the Q156 whole-table audit and re-verified here against a 200 kb query before correcting. |
| absence class | measured-not-analysed |
| epigenetic readout | none published. This beat's KMT2A CUT&RUN scoring in GSE252378 is the only measurement of any kind: 0 of 27 samples at the project's window, 23 of 27 at a 400 bp window on the 86,936,204 TSS. |
| shown for fusion | measured in KMT2A::AFF1, ::MLLT1, ::MLLT3, ::MLLT10, ::AFDN and ::EPS15 in GSE252378; the near-AFF1 window is occupied in 23 of 27 samples across ALL SIX partners, so this is not partner-specific |
gene synonyms
[ "MGC26744" ]
current curated tss hg38
[ 86892422, 86934732, 86936204 ]
transcripts per tss — the numbers behind the chart
{
"86892422": 13,
"86934732": 4,
"86936204": 3
}distance to aff1 tss bp — the numbers behind the chart
{
"86892422": 42579,
"86934732": 269,
"86936204": 1203
}two ensembl models same symbol
{
"short_model": {
"ensembl": "ENSG00000163633",
"span": "chr4:86876205-86892813",
"backed_by": "HGNC"
},
"long_model": {
"ensembl": "ENSG00000285458",
"span": "chr4:86886472-86936178",
"backed_by": "NCBI Gene 132989, GENCODE"
},
"tss_separation_bp": 43365
}pubmed bound
{
"C4orf36[All Fields]": 3,
"C4orf36[All Fields] AND leukemia[All Fields]": 0,
"C4orf36[All Fields] AND AFF1[All Fields]": 1,
"note": "the leukemia term expands through MeSH so the zero is against a wide net and is conservative; the one AFF1 co-hit is a TWAS where both are locus labels",
"date": "2026-09-24"
}measurements — the numbers behind the chart
[
{
"quantity": "tss_span",
"subject": "distance between C4orf36's most distal and most proximal current curated TSSs",
"value": 43782,
"units": "bp",
"n": 3,
"universe": "UCSC hg38 ncbiRefSeqCurated, chr4:86,800,000-87,000,000, 20 minus-strand transcripts at TSSs 86,892,422 / 86,934,732 / 86,936,204",
"null": 2000,
"note": "the null is the width of a promoter window: the gene's own TSSs are 22 window-widths apart and the table records one"
},
{
"quantity": "tss_distance_to_aff1",
"subject": "nearest current-RefSeq C4orf36 TSS to AFF1's TSS, opposite strands",
"value": 269,
"units": "bp",
"n": 2,
"universe": "ncbiRefSeqCurated NM_001414641.1 tss 86,934,732 against NM_001166693.3 tss 86,935,010, hg38",
"null": null,
"note": "inside the project's 2 kb AFF1 window; an equivalent 2 kb C4orf36 window would share 86.1% of it"
},
{
"quantity": "d_nearest",
"subject": "the project's C4orf36 row TSS against the symbol's nearest current curated TSS",
"value": 0,
"units": "bp",
"n": 1,
"universe": "promoters.tsv tss 86,892,422 against the three current TSSs; verified against UCSC ncbiRefSeqCurated over a 200 kb window",
"null": null,
"note": "CORRECTION. This entry first reported 42,310 bp, computed against a 30 kb track query that excluded the matching transcript."
},
{
"quantity": "n_samples_occupied",
"subject": "the project's C4orf36 promoter row (chr4:86,891,422-86,893,422), KMT2A-N1+N2, GSE252378",
"value": 0,
"units": "samples",
"n": 27,
"universe": "27 KMT2A-fusion samples, rule >= 2.0x in both KMT2A-N antibodies AND above the per-library max of eight tissue-restricted loci",
"null": 2.733,
"note": "null is the mean number of samples calling a promoter across all 19,404 rows under the same rule"
},
{
"quantity": "n_samples_occupied",
"subject": "a 400 bp window on the 86,936,204 C4orf36 TSS (chr4:86,936,001-86,936,401), KMT2A-N1+N2, GSE252378",
"value": 23,
"units": "samples",
"n": 27,
"universe": "same 27 samples, same rule, floor recomputed at 400 bp width over the same eight loci",
"null": null,
"note": "N mean 15.24, C mean 4.42, N/C 3.45 - the highest N/C anywhere in the 6 kb around AFF1, and outside the project's AFF1 window"
}
]- If any intervention is designed against the region immediately 5' of AFF1 - a CRISPRi guide, a degron-recruited effector, an oligonucleotide - it will engage a protein-coding gene's annotated promoter as well as AFF1's and an antisense RNA's, and the protein-coding one is uncharacterised, so its loss-of-function phenotype is unpredictable from the literature.
- More generally: a promoter table keyed on one TSS per symbol will report a clean negative for a gene that is present, current and hot at a different one of its own TSSs, and the negative is indistinguishable from a real one. Any target list filtered by 'no other gene's promoter nearby' inherits that failure.
- Are the 86,934,732 and 86,936,204 transcripts independently initiated, or are they 5' extensions and readthrough of the gene that starts at 86,892,422? Seventeen of the twenty curated transcripts belong to one recent accession batch, which is the signature of an automated extension. A 5'-end assay (CAGE, GRO-cap, 5' RACE) is the only thing that decides it, and everything this entry says about a C4orf36 promoter beside AFF1 depends on the answer.
- Is C4orf36 transcribed at all in B-lineage cells? Its three PubMed records are liver, bone marrow stromal and trophoblast, and no haematopoietic expression measurement was found.
- How should a one-row-per-symbol promoter table represent a gene with three TSSs 43 kb apart? Q156 measures 2,851 such symbols (14.69%) with a median span of 21,291 bp. Keeping the 5'-most, keeping all, or keeping the most-supported are three different tables and nobody has compared them.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
C4orf36 is NCBI GeneID 132989, HGNC:28386, 4q21.3, locus type 'gene with protein product', with HGNC refseq_accession NM_144645, HGNC ensembl_gene_id ENSG00000163633 and UniProt Q96KX1; the reviewed UniProt record CD036_HUMAN is 'Uncharacterized protein C4orf36', 117 aa.
HGNC record for C4orf36 (HGNC:28386) and UniProtKB search gene_exact:C4orf36 AND organism_id:9606 · HUGO Gene Nomenclature Committee; UniProt Consortium · HGNC / UniProt 2026What this evidence cannot show
Database records fetched 2026-09-24. The UniProt query returned four records; three are unreviewed TrEMBL fragments at 81, 113 and 46 aa and only Q96KX1 is reviewed, so the primary gene name was read back and the length asserted per the standing accession rule. A reviewed accession for an uncharacterised ORF says the sequence is curated, not that a protein has ever been detected.
Evidence that argues against this (1)
- HGNC's ensembl_gene_id for this symbol, ENSG00000163633, is the SHORT model that stops 42 kb from AFF1. So the nomenclature authority's own cross-reference points away from the end of the gene that matters here, and a reader taking HGNC as the arbiter would conclude C4orf36 is nowhere near AFF1. Ensembl REST API et al. 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
In the current UCSC hg38 ncbiRefSeqCurated track over chr4:86,800,000-87,000,000, C4orf36 has twenty curated minus-strand transcripts at exactly three distinct TSSs - 86,892,422 (13 transcripts including NM_144645.4), 86,934,732 (NM_001414641.1, NM_001414642.1, NR_183021.1, NR_183022.1) and 86,936,204 (NM_001414639.1, NM_001414640.1, NR_183020.1) - spanning 43,782 bp, so a C4orf36 TSS lies 269 bp from AFF1's TSS inside the project's 2 kb AFF1 window.
UCSC Genome Browser hg38 ncbiRefSeqCurated track, chr4:86,800,000-87,000,000 · UCSC Genome Browser, api.genome.ucsc.edu · UCSC Genome Browser 2026What this evidence cannot show
This query deliberately spans 200 kb because the FIRST version of this entry used a 30 kb window (chr4:86,915,000-86,945,000) that excluded the 86,892,422 TSS and therefore produced a false displacement claim. Coordinates are 0-based half-open txStart / 1-based txEnd; for a minus-strand transcript the TSS is txEnd. Whether any of these transcripts is expressed in a haematopoietic cell is unknown and is not claimed; an annotated TSS is a transcript-model start, not a demonstrated initiation site.
Evidence that argues against this (1)
- Three annotated TSSs for one uncharacterised ORF, seventeen of whose twenty transcripts were added in a single recent RefSeq batch (the NM_0014146xx / NR_1830xx series), is the signature of an automated 5' extension rather than of three curated promoters. It is also possible the long transcripts are readthrough products, in which case there is no C4orf36 promoter at 86,934,732 or 86,936,204 at all. Nothing in the annotation distinguishes those cases; CAGE or GRO-cap would. Core LJ et al. 2008
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
The project's C4orf36 promoter row at chr4:86,891,422-86,893,422, keyed on NM_144645 and derived from a 2020 refGene snapshot, sits on one of the three current TSSs and is NOT displaced (d_nearest = 0); a whole-table audit of all 19,404 rows against the current ncbiRefSeqCurated finds 81.63% TSS-identical, 1.83% genuinely displaced and 14.69% of symbols carrying multiple non-overlapping TSSs with a median span of 21,291 bp.
The promoter table's second blind spot is not annotation drift at 1.8% - it is one row per symbol at 14.7%, and the case that motivated the question is not displaced at all · Iqbal Lab oncofusion knowledge graph, beat Q156 (coordinator) · this project 2026What this evidence cannot show
Q156's audit is cited here as the source that CORRECTED this entry, and its C4orf36 result was independently re-checked by this beat against the UCSC ncbiRefSeqCurated track over a 200 kb window before the correction was applied - it is not taken on report. Q156's whole-table figures (81.63%, 1.83%, 14.69%) are quoted and were NOT re-derived here; they should be re-run before being used as load-bearing quantities elsewhere.
Evidence that argues against this (1)
- That the row is not displaced makes the situation WORSE for the project, not better. A displaced row could be fixed by refreshing the annotation. A symbol with three legitimate TSSs and a one-row-per-symbol schema cannot be fixed that way at all, and Q156 measures that class at eight times the displacement rate. UCSC Genome Browser et al. 2026
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
Ensembl carries two distinct protein-coding genes under the symbol C4orf36: ENSG00000163633 at chr4:86,876,205-86,892,813 sourced from HGNC, and ENSG00000285458 at chr4:86,886,472-86,936,178 sourced from NCBI Gene 132989, both on the minus strand, so their TSSs differ by 43,365 bp.
Ensembl REST lookup/id for ENSG00000163633 and ENSG00000285458, and overlap/region/human/4:86915000-86945000, GRCh38 · Ensembl REST API, rest.ensembl.org · Ensembl 2026What this evidence cannot show
Ensembl REST returned HTTP 500 and 503 several times during this beat; both lookups here are responses that were received and saved. Two gene ids sharing a display name is normal in Ensembl when a locus has both an HGNC-anchored and an NCBI-anchored model, and it does not mean Ensembl believes there are two genes - but it does mean a symbol-keyed lookup can return either, with no error and no warning. Note this is the Ensembl-side counterpart of the three-TSS RefSeq picture, not an independent observation of it.
Evidence that argues against this (1)
- The UCSC knownGene (GENCODE) track labels BOTH ENST00000473559 (TSS 86,934,732) and ENST00000503001 (TSS 86,936,178) as C4orf36, and annotates the second as protein_coding_CDS_not_defined - i.e. GENCODE itself does not commit to the long model encoding the protein. Treating the long model as simply 'the C4orf36 gene' overstates the annotation's confidence. UCSC Genome Browser et al. 2026
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
In GSE252378, the project's C4orf36 promoter row (chr4:86,891,422-86,893,422) is occupied in 0 of 27 samples, while a 400 bp window centred on the 86,936,204 C4orf36 TSS is occupied in 23 of 27 and has a KMT2A-N mean of 15.24 against a KMT2A-C mean of 4.42.
Q155 per-sample peak scoring and promoter-matrix occupancy at the AFF1/C4orf36 locus, GSE252378 · Iqbal Lab oncofusion knowledge graph, beat Q155 · this project 2026What this evidence cannot show
No spike-in exists in this deposit, so all values are within-sample shares. The 0 of 27 and the 23 of 27 are NOT a like-for-like comparison of the same statistic: the first is a 2,000 bp window scored against an 8-locus floor at 2 kb width, the second a 400 bp window scored against the same 8 loci at 400 bp width. Both use the same rule and floor construction and the direction of the difference is far larger than the width effect, but the two numbers should not be subtracted. Crucially, the 23 of 27 is NOT partner-specific - 7 of 8 KMT2A::AFF1 samples and 16 of 19 others - so it is a property of the locus, not of the fusion's own partner.
Evidence that argues against this (1)
- An IgG hotspot sits 150-250 bp from that window: the IgG arm's own maximum in the whole 6 kb is at chr4:86,936,400-86,936,450 at 11.86, and a 400 bp window there is occupied in 20 of 27 samples with an N/C median of 2.53 - HIGHER than the peak's 2.32. So this stretch of chromosome is generically sticky and the N/C ratio does not by itself distinguish a fusion peak from an artefact here. Iqbal Lab oncofusion knowledge graph et al. 2026
C4orf36 is a multi-TSS symbol, not a displaced row
current-RefSeq TSS 269 bp from AFF1's on the opposite strand, inside the project's 2 kb AFF1 window
the third TSS in the same 2 kb window; C4orf36's body spans AFF1-AS1's entirely under the long model
the second way a neighbour goes missing: present in the table, placed 42,310 bp away by an annotation snapshot
the third gene, found as a by-product of a control run for something else
and 86.10% of it is C4orf36's under current annotation
the global maximum of the 6 kb interval sits on C4orf36's current-RefSeq long TSS, 249 bp outside the project's AFF1 window
the worked case for displacement and for one-symbol-two-gene-models
d_nearest = 0: the project's C4orf36 row sits on one of the symbol's three current TSSs and is not displaced