IRX2-DT (C5orf38) - the divergent partner 215 bp from the IRX2 TSS that makes every 2 kb IRX2 window ambiguous
also called C5orf38 · CEI · IRX2NB · IRX2 divergent transcript · IRX2 neighbor
Right next to the IRX2 gene, pointing the opposite way, sits another gene that shares the same on-switch. Any measurement of that switch is a measurement of both genes at once, so a result at IRX2 cannot be assigned to IRX2 alone.
The same thing, for a scientist
IRX2-DT, NCBI GeneID **153571**, previous official symbol **C5orf38**, 5p15.33, chr5:2,751,892-2,758,577 on the **plus** strand. Its TSS lies **215 bp** from IRX2's on the **opposite** strand - a canonical divergent promoter pair - so the two genes' 2 kb TSS windows overlap by **1,785 bp, 89.2%** of each [E1][E2]. NCBI's own other-designations for it are *"coordinated expression to IRX2"* and *"IRX2 neighbor"*, i.e. the gene is named for being co-expressed with IRX2 [E1]. Despite the *-DT* (divergent transcript) symbol it is **not purely non-coding**: hg38 refGene lists **seven** protein-coding transcripts with a CDS from 2,752,264 and two non-coding ones (this sentence said "eight" until cycle 139; the entry's own data.refgene_transcripts_coding field listed seven and refGene, re-fetched by beat Q153 over chr5:2,735,000-2,765,000, gives seven) [E2]. This matters to this project for one concrete reason. Every IRX2 promoter measurement in the q145/q147/q151 machinery keys on a 2 kb window that is 89.2% IRX2-DT, and in the single sample where that window carries signal (p454, KMT2A::AFF1) the IRX2-DT row reads **3.063 / 1.886** against IRX2's **3.812 / 1.898** in the same two libraries - the same signal, twice [E3]. IRX2-DT is occupied in **0 of 27** samples under the same rule, as is IRX2 [E3]. This graph had no entry for it and the corpus names it nowhere.
Coordinates, fetched not remembered
UCSC hg38 refGene via api.genome.ucsc.edu/getData/track, fetched 2026-09-24, chr5:2,745,000-2,760,000:
| gene | transcript | interval | strand | CDS |
|---|---|---|---|---|
| IRX2 | NM_001134222 | chr5:2,745,844-2,751,677 | - | 2,747,563-2,751,413 |
| IRX2 | NM_033267 (MANE) | chr5:2,746,164-2,751,677 | - | 2,747,563-2,751,413 |
| C5orf38 / IRX2-DT | NM_001306149 | chr5:2,751,892-2,758,578 | + | 2,752,264-2,753,359 |
| C5orf38 / IRX2-DT | NM_178569, NM_001294337 | chr5:2,752,130-2,755,397 | + | 2,752,264-2,755,041 |
| C5orf38 / IRX2-DT | NR_122125, NR_122126 | chr5:2,751,892-2,758,578 | + | non-coding |
NCBI esummary GeneID 153571 returns chrstart 2751892, chrstop 2758577, maplocation 5p15.33, aliases C5orf38, CEI, IRX2NB.
TSS separation from IRX2: 2,751,892 - 2,751,677 = 215 bp, head to head. 12
What this does to every IRX2 window this project uses
The project's promoter table (C:\Temp\q145irx1\promoters.tsv, built from refGene and strand-aware) gives:
IRX2 chr5 2751677 2750677 2752677 - NM_001134222
C5orf38 chr5 2751892 2750892 2752892 + NM_001306149
Overlap = 2,752,677 - 2,750,892 = 1,785 bp of 2,000, i.e. 89.2%.
So the two rows are near-duplicate measurements. In p454, the only GSE252378 sample where either carries appreciable signal:
| library | IRX2 share | IRX2-DT share |
|---|---|---|
| p454 KMT2A-N1 | 3.812 | 3.063 |
| p454 KMT2A-N2 | 1.898 | 1.886 |
| p454 KMT2A-C1 | 1.595 | 1.595 |
| p454 KMT2A-C2 | 1.106 | 1.106 |
| p454 IgG | 0.000 | 0.000 |
The C-side arms are identical to four decimal places, which is what two 89%-overlapping means of the same coverage track should look like. Neither clears the occupancy rule in any of the 27 samples. 3
The consequence is not "add a caveat". It is that a 2 kb mean cannot answer which promoter, and only base-resolution signal from the bigWig could - which is a different measurement from the one this project's matrix supports.
What is not known about it
Almost everything. There is no expression measurement of IRX2-DT in any KMT2A-rearranged cohort in this graph, and the q149 GPL570 panel does not include a probe for it. C5orf38[tiab] returns 7 PubMed records and IRX2-DT[tiab] returns 0 (querytranslation IRX2-DT[tiab], untranslated, so that zero is weaker than it looks - a hyphenated symbol is a query hazard in this index and the zero is reported as such rather than as a bound). Nothing in the corpus names it.
It is written here because it is the limit on a measurement this project has now made, not because there is a story about it.
| ncbi gene ID | 153571 |
|---|---|
| gene primary | IRX2-DT |
| previous official symbol | C5orf38 |
| description | IRX2 divergent transcript |
| cytoband | 5p15.33 |
| uniprot | Q86SI9 |
| divergent pair with | gene-irx2 |
| tss separation from irx2 bp | 215 |
| promoter window overlap with irx2 bp | 1785 |
| promoter window overlap with irx2 PCT | 89.2 |
| promoter window used by this project | chr5:2750892-2752892 (TSS 2751892 +/- 1000, keyed on NM_001306149) |
| epigenetic readout | none recorded - no histone mark, accessibility or methylation measurement at IRX2-DT exists in this corpus; the only measurement of any kind is this beat's KMT2A occupancy scoring |
| shown for fusion | measured in KMT2A::AFF1, ::MLLT1, ::MLLT3, ::MLLT10, ::AFDN and ::EPS15 in GSE252378 and NOT occupied in any of the 27 samples; nobody has looked at expression in any fusion |
| occupancy gse252378 | 0 of 27 samples under the q147 two-antibody 2.0x plus width-matched-floor rule |
| length AA | 138 |
| uniprot entry name | CEI_HUMAN |
| protein name | Putative uncharacterized protein IRX2-DT |
| resolved how | UniProt REST search gene:C5orf38 AND organism_id:9606, fetched 2026-09-24 by beat Q153; ONE reviewed record; primary gene name read back as IRX2-DT (synonyms C5orf38, CEI) and length asserted so verify-accessions.py can catch a swap. An unreviewed B7ZL86 "C5orf38 protein" at 135 aa is also returned and is NOT used. UniProt calls the product "putative uncharacterized", so the accession settles the identifier and settles nothing about whether a protein is made - Ensembl annotates the same locus as a lincRNA (ENSG00000186493, logic_name ensembl_havana_lincrna). |
| uniprot note withdrawn cycle 139 | WITHDRAWN 2026-09-24 by beat Q153, which looked the accession up. This field previously read: "no reviewed human UniProt accession was resolved for this locus in this beat; refGene lists protein-coding transcripts with a CDS from chr5:2,752,264 but the accession was NOT looked up, and it is recorded absent rather than guessed" |
gene synonyms
[ "C5orf38", "CEI", "IRX2NB" ]
other designations
[ "IRX2 neighbor", "coordinated expression to IRX2", "coordinated expression to IRXA2 homeobox" ]
hg38
{
"ncbi_esummary": "chr5:2751892-2758577, plus strand (chrstart < chrstop)",
"refgene_earliest_tss": 2751892,
"refgene_transcripts_coding": [
"NM_001365689",
"NM_001365688",
"NM_001365685",
"NM_001306149",
"NM_001306150",
"NM_001294337",
"NM_178569"
],
"refgene_transcripts_noncoding": [
"NR_122125",
"NR_122126"
],
"strand": "+"
}pubmed bound
{
"C5orf38[tiab]": 7,
"IRX2-DT[tiab]": 0,
"note": "the IRX2-DT zero has querytranslation 'IRX2-DT[tiab]' - untranslated - so a hyphenated-symbol parse hazard cannot be excluded and this is NOT recorded as a bounded absence",
"date": "2026-09-24"
}measurements — the numbers behind the chart
[
{
"quantity": "tss_separation",
"subject": "IRX2-DT TSS relative to IRX2 TSS, hg38",
"value": 215,
"units": "bp",
"n": 1,
"universe": "UCSC hg38 refGene: IRX2 NM_001134222 minus strand TSS 2,751,677; C5orf38 NM_001306149 plus strand TSS 2,751,892",
"null": null,
"note": "head to head, opposite strands - a canonical divergent promoter pair"
},
{
"quantity": "promoter_window_overlap",
"subject": "IRX2 and IRX2-DT 2 kb TSS windows in this project's promoter table",
"value": 89.2,
"units": "percent of a 2,000 bp window",
"n": 2,
"universe": "IRX2 chr5:2,750,677-2,752,677 against C5orf38 chr5:2,750,892-2,752,892; overlap 1,785 bp",
"null": null,
"note": "this is why no 2 kb call at this locus can name its gene"
},
{
"quantity": "n_samples_occupied",
"subject": "IRX2-DT 2 kb promoter, KMT2A-N1+N2, GSE252378",
"value": 0,
"units": "samples",
"n": 27,
"universe": "27 KMT2A-fusion samples, rule >=2.0x both KMT2A-N antibodies AND above the 8-locus width-matched floor",
"null": 2.73,
"note": "same null as every other promoter in this deposit; identical result to IRX2, as 89% shared sequence requires"
}
]- Any intervention aimed at the IRX2 promoter necessarily engages IRX2-DT's promoter as well - they are 215 bp apart and share their regulatory sequence - so a selective IRX2 perturbation at the DNA level is not available, and a phenotype from such a perturbation cannot be attributed to IRX2 alone.
- Is IRX2-DT expressed in KMT2A::AFF1 infant ALL? GPL570's probe set as used by q149 contains no IRX2-DT probe, and no RNA-seq cohort in this graph has been queried for it.
- ANSWERED in cycle 139 by beat Q153: yes - Q86SI9, CEI_HUMAN, 138 aa, reviewed Swiss-Prot, primary gene name IRX2-DT. Original question: "Does IRX2-DT have a reviewed UniProt accession? Not looked up in this beat and recorded absent rather than guessed; refGene's CDS says it should."
- In the one sample with signal (p454), does base-resolution bigWig coverage place the summit over IRX2's TSS or IRX2-DT's? A 215 bp question that a 2 kb mean cannot answer.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
IRX2-DT (NCBI GeneID 153571, previous official symbol C5orf38, aliases CEI and IRX2NB) is at 5p15.33, chr5:2,751,892-2,758,577, and NCBI's other-designations for it include 'coordinated expression to IRX2' and 'IRX2 neighbor'.
IRX2-DT IRX2 divergent transcript [Homo sapiens (human)], Gene ID 153571 · NCBI Gene · NCBI Gene 2026What this evidence cannot show
A database record fetched 2026-09-24 via esummary and efetch, not a paper. NCBI coordinates are 0-based half-open in esummary's chrstart/chrstop; the 1-based TSS used elsewhere in this entry comes from UCSC refGene. The 'coordinated expression to IRX2' designation is a naming convention, not a measurement of co-expression in any leukaemia.
Evidence that argues against this (1)
- hg38 refGene lists seven NM_ (protein-coding) transcripts for this locus with a CDS from 2,752,264, alongside two NR_ ones, and UniProt carries a reviewed accession Q86SI9 (CEI_HUMAN, 138 aa). Treating it as purely non-coding would be wrong - although Ensembl DOES annotate it as a lincRNA (ENSG00000186493, logic_name ensembl_havana_lincrna) and UniProt calls the product 'putative uncharacterized', so the two major annotations disagree about whether this gene codes. (This note said 'eight' until cycle 139; corrected by beat Q153 against refGene re-fetched over chr5:2,735,000-2,765,000.) UCSC Genome Browser et al. 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
In hg38 refGene, IRX2-DT/C5orf38 transcribes from the plus strand with its earliest TSS at chr5:2,751,892 while IRX2 transcribes from the minus strand with its TSS at chr5:2,751,677, a head-to-head separation of 215 bp, so their 2 kb TSS windows overlap by 1,785 bp (89.2%).
UCSC Genome Browser hg38 refGene track, chr5:2,745,000-2,760,000 · UCSC Genome Browser, api.genome.ucsc.edu · UCSC Genome Browser 2026What this evidence cannot show
refGene is one annotation among several; Ensembl and GENCODE place transcript starts slightly differently, and Ensembl REST returned HTTP 500 during this beat so it was not cross-checked. The 89.2% figure is specific to the 2 kb window convention used by this project's promoter table; a 500 bp window would overlap less and a 5 kb window more. The UCSC HGNC track places IRX2-DT at 2,752,131, which is the start of the NM_178569/NM_001294337 transcripts rather than the earliest one.
Evidence that argues against this (1)
- Divergent promoter pairs are common in mammalian genomes, so this geometry is not special to IRX2 and does not by itself imply shared regulation. What it does imply is a measurement limit, which is the only claim made here. NCBI Gene 2026
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
In GSE252378 the IRX2-DT/C5orf38 2 kb promoter is occupied in 0 of 27 KMT2A-fusion samples under the same two-antibody rule as IRX2, and in p454 - the only sample with appreciable signal at this locus - it reads 3.063 and 1.886 against IRX2's 3.812 and 1.898 in the same two KMT2A-N libraries, with the two C-side arms identical to four decimal places.
q151 IRX2 occupancy scoring and divergent-transcript decomposition, GSE252378 · Iqbal Lab oncofusion knowledge graph, beat Q151 · this project 2026What this evidence cannot show
The two rows are 89.2% the same sequence, so their agreement is arithmetic rather than corroboration and is reported as such. No spike-in exists in this deposit so all values are within-sample shares. n = 1 sample carries the comparison.
Evidence that argues against this (1)
- Because the windows overlap so heavily, this comparison CANNOT distinguish a peak centred on IRX2's TSS from one centred on IRX2-DT's. The small IRX2-over-IRX2-DT excess in KMT2A-N1 (3.812 vs 3.063) is consistent with a peak slightly nearer IRX2's TSS but is well within what a 215 bp offset between two 2 kb means would produce either way. UCSC Genome Browser et al. 2026
- co-occurs with →IRX2 - the IRX-cluster neighbour that is this disease's own prognostic gene, activated only in KMT2A::AFF1, and never written here until now
divergent promoter pair: TSSs 215 bp apart on opposite strands, so their 2 kb windows overlap by 1,785 bp (89.2%)
- related to →IRX2 is not fusion-occupied at promoter scale in GSE252378: 0 of 27 samples, against a null of 2.73
makes every 2 kb IRX2 promoter measurement in this project ambiguous between the two genes
- related to →IRX1 — Iroquois homeobox 1, the gene that names the HOXA-negative half of infant t(4;11) ALL
both sit in the 5p15.33 IRX cluster; IRX1 is plus strand and IRX2 minus, which is why the window recipe does not transfer
the IRX2/IRX2-DT 89.2% window overlap is the precedent this check exists for
re-derives the 215 bp / 89.2% geometry from three fetched annotations and confirms it; supplies the UniProt accession Q86SI9 the entry records as absent, and corrects 'eight' coding transcripts to seven
- ← co-occurs withIRX2 - the IRX-cluster neighbour that is this disease's own prognostic gene, activated only in KMT2A::AFF1, and never written here until now
head-to-head divergent pair, TSSs 215 bp apart on opposite strands, gene bodies disjoint
the same 27 zeros are a bounded negative for IRX2-DT as well - the one dividend a shared window pays
the same divergent-neighbour problem one step worse: 151 bp vs 215 bp, 92.45% vs 89.2%, and neither gene has a row in the project's promoter table
the same three-annotation disagreement Q153 recorded for IRX2/IRX2-DT, one order of magnitude wider here