Oncofusion

start / graph / absence-bard1-and-exo1-are-the-only-two-resection-proteins-with-no-occupancy-data-in-any-archive

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q37-archive-sweep

BARD1 and EXO1 are the only two proteins in the resection layer with ZERO chromatin-occupancy datasets anywhere - ENCODE, ChIP-Atlas, ReMap and a hand-checked GEO title list all return nothing - and BARD1's silence is the loud one, because its obligate partner BRCA1 has thirty-one

also called the two proteins with no occupancy data anywhere · BARD1 has no ChIP and BRCA1 has thirty-one · EXO1 in GEO means exosome

Of the twelve DNA-repair proteins this graph tracks around the KMT2A gene, ten have been mapped onto chromatin somewhere in a public dataset. Two have never been mapped at all, and one of them is the obligate partner of a protein that has been mapped thirty-one times, which points at a missing reagent rather than a missing idea.

The same thing, for a scientist

**This is the residue of the cycle-62 six-protein absence after three archives were swept** [E1, E2]. Q32 classed BARD1, EXO1 and DNA2 as `never-measured` on the basis of an ENCODE HTTP 404, and Q35 kept that classification. Two of the three do not survive: DNA2 has ChIP-seq in two independent GEO series, and KDM5A - also carried as never-measured in a blood cell - has ChIP-seq in NB-4 and MM.1S. **BARD1 and EXO1 do survive, and they survive every check.** | check | BARD1 | EXO1 | |---|---|---| | ENCODE released experiments, all assay types | **2** (ENCSR104RTM, ENCSR427KVG) - **both CRISPR RNA-seq** | 0 | | ENCODE target record | **YES**, `/targets/BARD1-human/` HTTP 200 | no, HTTP 404 | | ENCODE ChIP-seq / CUT&RUN / any occupancy assay | **0** | **0** | | ChIP-Atlas distinct SRX (hg19+hg38) | **0** | **0** | | ReMap 2022 datasets | **0** (HTTP 401 "no experiment for target BARD1") | **0** | | GEO GSMs with the symbol in the title | 13 | 55 | | of which are an occupancy assay | **0** (all RNA-Seq, PAR-CLIP, easyCLIP, 3'READS) | **0** | **The GEO title lists were read by hand rather than counted, because at n = 13 and n = 55 a count is not a result** [E2]. BARD1's thirteen are ovarian-line siRNA and overexpression RNA-seq, a PAR-CLIP, two easyCLIP replicates, three A673 samples whose `library_strategy` reads `RNA-Seq`, and one 3'READS/TREND-seq sample. **EXO1's fifty-five are dominated by a collision:** in GEO sample titles `exo1` overwhelmingly means *exosome preparation 1* - `hUCMSC-EXOs exo1`, `MSC_Exo1`, `ALCL plasma S-EVs REC_exo1`, `Cavin-1PC3_Exo1`, `HUVEC+HeLa exo_exo1` - and the remainder are EXO1-knockout editing-outcome libraries. Not one is chromatin. **Why BARD1's zero is the informative one.** BARD1 and BRCA1 are an obligate heterodimer, they are co-recruited in the model this graph carries, their DepMap profiles are near-identical (**BARD1 -0.574 with 58.2% of 1,258 lines below -0.5, BRCA1 -0.489 with 44.0%**) and they covary across cell lines at **r = 0.560**. **And BRCA1 has 31 ChIP-Atlas experiments in 12 cell types while BARD1 has none.** A 31-to-0 asymmetry between two halves of one obligate dimer is not a statement about how interesting BARD1 is. **The likeliest explanation is that no ChIP-grade BARD1 antibody has been validated**, and that is a checkable, reagent-level fact this beat did not check [E3]. **EXO1's zero is differently informative and arguably not a gap at all.** EXO1 is a processive 5'-3' exonuclease that acts on resected ssDNA, transiently and sub-stoichiometrically. **ChIP is close to the wrong assay in principle** - which is what Q32 said and was right about - and the right instruments (END-seq, BLESS, ssDNA-resection qPCR, and the DNA-PK-bound-fragment sequencing of GSE218590) do not produce a "EXO1 ChIP-seq" record for any archive to index. **So EXO1's `never-measured` should be read as `no-suitable-assay-has-been-pointed-here`, not as neglect.** **Absence class, per protein, stated for quoting:** BARD1 `never-measured`, and the cost of closing it is an antibody before it is an experiment. EXO1 `never-measured` by ChIP and `never-measured` at *KMT2A* by any resection assay, because no break has been induced there (see `ana-q37-a-damage-state-resection-occupancy-dataset-does-exist-and-it-cannot-reach-the-kmt2a-locus`).

Facts and figures
scriptanalyses/q37-archive-sweep-at-the-kmt2a-locus/run.py and q37_geo.py
statusrun
run date2026-09-22
run byQ37-archive-sweep
absence classnever-measured
absence class noteThis is the ONE entry in this beat where never-measured is the right headline class, and it applies to exactly two of the twelve proteins censused. Every other protein in that census has occupancy data somewhere.
what is absentAny measurement of BARD1 or EXO1 occupancy on chromatin, at any locus, in any human cell, in ENCODE, ChIP-Atlas, ReMap 2022 or the GEO sample titles carrying their symbols.
what is NOT absent(1) BARD1 RNA-level perturbation data - two ENCODE CRISPR RNA-seq experiments and several GEO siRNA/overexpression series. (2) BARD1 RNA-binding data - a PAR-CLIP and two easyCLIP replicates. (3) EXO1 knockout editing-outcome libraries. (4) Dependency data for both - BARD1 -0.574 and EXO1 -0.127 over 1,258 DepMap screens each, Open Targets 26.06. None of these is occupancy.
the encode 404 correctionQ32 and Q35 both record 'BARD1: no ENCODE target record exists (HTTP 404)'. That is false. /targets/?label=BARD1 returns HTTP 200 with /targets/BARD1-human/, and the experiment search returns 2. The 404 they saw came from adding assay_title=TF ChIP-seq. An ENCODE 404 is a property of the whole query. RAD50 is the clean demonstration: target record HTTP 200, experiment search HTTP 404.
how remap spells zeroHTTP 401 with body {"error":"Sorry. There is no experiment for target BARD1 and tax_id 9606 on remap2022"} - not 404, not an empty 200. CTCF returned 628 datasets in the same batch.
relationship to existing entriesNarrows `absence-no-resection-factor-localised-at-the-kmt2a-locus` from six never-measured proteins to two, and supplies the archive checks that entry names as not run. It does not restate its 21 PubMed queries, which stand.
absence class per protein
{
  "BARD1": "never-measured (chromatin occupancy). ENCODE: a target record and two CRISPR RNA-seq experiments, zero ChIP. ChIP-Atlas 0. ReMap 0. GEO: 13 title GSMs, 0 occupancy, hand-checked. What closes it: a validated ChIP-grade antibody, then an experiment. The reagent is the bottleneck, not the question.",
  "EXO1": "never-measured (chromatin occupancy), and ChIP is arguably the wrong assay class. ENCODE: no target record. ChIP-Atlas 0. ReMap 0. GEO: 55 title GSMs, 0 occupancy, dominated by an exosome naming collision. What closes it: END-seq, BLESS or ssDNA-resection qPCR at an induced break, not a ChIP."
}
geo hand check
{
  "instrument": "esearch db=gds, <SYMBOL>[Title] AND \"Homo sapiens\"[Organism] AND gsm[ETYP], querytranslation read back, then esummary titles read one by one",
  "BARD1_13_titles": [
    "SKOV3ip1 BARD1 siRNA_1/2/3 (RNA-Seq)",
    "OVCAR8ip1 BARD1 expressed_1/2/3 (RNA-Seq)",
    "BARD1 PAR-Clip",
    "A673 BARD1 rep1/2/3 (library_strategy RNA-Seq, GSE182677)",
    "easyCLIP BARD1 Replicate 1/2 MiSeq",
    "BARD1 (1_013) (3'READS / TREND-seq)"
  ],
  "EXO1_collision": "of 55, the majority use exo1 to mean an exosome preparation. Examples read: hUCMSC-EXOs exo1 [miRNA-seq], CAFs cells exo1, GBM39 cells Exo1, MSC_Exo1, ALCL plasma S-EVs REC_exo1 and CR_exo1, SPRI/Qiam Exo1 Treatment, S-patient-exo1, shMCU-exo1, Cavin-1PC3_Exo1, GFPPC3_Exo1, HUVEC+HeLa exo_exo1. Two GSMs titled bare 'EXO1' were opened: GSM8598015 library_strategy RNA-Seq, GSM4296141 library_strategy ncRNA-Seq.",
  "assay_token_note": "a GEO query cannot search the literal string CUT&RUN, because Entrez parses the ampersand as the boolean AND: CUT&RUN[All Fields] translates to CUT[All Fields] AND RUN[All Fields]. That form is BROADER, not narrower, so the zeros above are conservative; it was read back from querytranslation rather than assumed."
}
bard1 versus brca1 asymmetry
{
  "BRCA1_chipatlas_SRX": 31,
  "BARD1_chipatlas_SRX": 0,
  "BRCA1_depmap_mean": -0.4888,
  "BARD1_depmap_mean": -0.5742,
  "BRCA1_BARD1_correlation_across_1208_lines": 0.56,
  "reading": "the two halves of an obligate heterodimer are equally strong dependencies and covary tightly, and one has 31 occupancy datasets and the other none. That asymmetry is about reagents or convention, not about the biology's interest.",
  "the_unchecked_check": "whether a ChIP-validated BARD1 antibody exists (Antibody Registry, ENCODE antibody characterisation documents, or the methods of any BRCA1 ChIP paper that also blots BARD1) was NOT checked by this beat and is named here as an open, cheap task rather than left implicit."
}
How this could bite us back
  • Declaring BARD1 never-measured makes the next beat propose an experiment. If the real obstacle is that no antibody works, that proposal is a reagent-development project wearing an experiment's clothes, and the honest cost estimate is an order of magnitude higher than the entry implies.
  • Declaring EXO1 never-measured invites a ChIP that would fail for principled reasons - a transient, sub-stoichiometric nuclease on resected single-stranded DNA is close to the worst case for crosslinking chromatin immunoprecipitation - and a failed ChIP would then be recorded somewhere as a negative result about EXO1 at this locus.
Still open
  1. Does a ChIP-validated BARD1 antibody exist? The Antibody Registry, the ENCODE antibody characterisation documents and the methods sections of the BRCA1 ChIP papers would settle it in under an hour, and the answer decides whether BARD1's absence costs an antibody-development project or a single afternoon of ChIP.
  2. Has EXO1-dependent resection ever been mapped genome-wide by any assay, under any name? The END-seq, BLESS, ssDNA-seq and DNA-PK-fragment literatures were not queried by name in this beat, only the ChIP-shaped archives, and rule 71 says a pathway is often named by its assay rather than by its enzyme.
  3. Is there any protein in this graph's whole interactome with the BARD1 signature - a well-studied obligate partner of a heavily ChIPped protein, itself never ChIPped? The census that would find them is the same three-archive sweep run over the 297 partner entries, and it has never been run.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    BARD1 and EXO1 have zero chromatin-occupancy datasets in all three archives checked on 2026-09-22. ENCODE: BARD1 has a target record and two released experiments, both CRISPR RNA-seq (ENCSR104RTM HepG2, ENCSR427KVG K562), and zero ChIP-seq; EXO1 has no target record at all. ChIP-Atlas: zero distinct SRX for either, from a live stream of experimentList.tab with both exact-match and alias-regex passes. ReMap 2022: zero for both, returned as HTTP 401 with an explicit 'no experiment for target' message, against CTCF's 628 in the same batch.

    PMID 42051034DOI
    What this evidence cannot show

    Three archives is not all archives: IHEC and BLUEPRINT were not queried directly (their TF ChIP-seq content is negligible and their histone data is irrelevant here, but that is an assumption), and PRIDE, 4DN and the Human Cell Atlas were not queried at all. ChIP-Atlas and ReMap both index SRA-deposited reads, so they are not independent of each other; a study deposited only as processed files is invisible to both. A zero across three overlapping archives is a stronger bound than a zero in one, but it is not a bound on the literature.

    Evidence that argues against this (1)
    • ReMap 2022 lists ENCSR857KDI (BRCA1, A549), which the live ENCODE released-experiment search does not return. So the archives disagree about what ENCODE contains, in the direction of ReMap holding MORE - which means an ENCODE-derived zero can be an undercount, and by extension so can a three-archive zero. Hammal F et al. 2022
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    Every GEO sample title carrying either symbol was read rather than counted, and none is an occupancy assay. BARD1 returns 13 GSMs: ovarian-line siRNA and overexpression RNA-Seq, a PAR-CLIP, two easyCLIP replicates, three A673 samples whose library_strategy field reads RNA-Seq, and one 3'READS/TREND-seq sample. EXO1 returns 55, of which the majority use 'exo1' to mean an exosome preparation (hUCMSC-EXOs exo1, MSC_Exo1, ALCL plasma S-EVs REC_exo1, Cavin-1PC3_Exo1, HUVEC+HeLa exo_exo1) and the rest are EXO1-knockout editing-outcome libraries; two GSMs titled bare 'EXO1' were opened and are RNA-Seq and ncRNA-Seq.

    What this evidence cannot show

    The query is scoped to the SYMBOL IN THE SAMPLE TITLE, which is the tightest and therefore the most under-recovering form. It demonstrably misses samples named by a protein name or a retired symbol: in this same sweep NBN[Title] returns 0 ChIP GSMs while NBS1[Title] returns 19, and RBBP8[Title] returns 0 while CtIP[Title] returns 16. For BARD1 the protein has no widely used alternative name and for EXO1 the alternatives (hExo1, Exonuclease 1) were NOT searched, which is the largest weakness of the EXO1 zero.

    Evidence that argues against this (1)
    • The NBN and RBBP8 cases in this beat prove the query form used here can return a confident zero for a protein that has nineteen datasets. BARD1 and EXO1 are claimed never-measured partly on this form, and the only defence is that ChIP-Atlas and ReMap index by antigen rather than by title and also return zero. Q37-archive-sweep 2026
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    BARD1's zero is anomalous against its own partner rather than against the field: BARD1 and BRCA1 are an obligate heterodimer with near-identical DepMap profiles (BARD1 mean -0.5742 with 58.2% of 1,258 lines below -0.5; BRCA1 -0.4888 with 44.0%) that covary across 1,208 shared lines at Pearson r = 0.560 (95% CI 0.520-0.598), and BRCA1 has 31 distinct ChIP-Atlas occupancy experiments in 12 cell types while BARD1 has none.

    What this evidence cannot show

    The inference from 31-versus-0 to 'no validated ChIP-grade antibody' is an inference and was NOT checked against the Antibody Registry, the ENCODE antibody characterisation documents, or the methods of any BRCA1 ChIP paper. It is the likeliest explanation among several - BARD1 may simply have been assumed redundant with BRCA1 by everyone who could have done the experiment - and it is recorded as a named, cheap, unexecuted check rather than as a finding.

    Evidence that argues against this (1)
    • An equally good explanation needs no reagent gap: if BARD1 is believed to be quantitatively co-recruited with BRCA1, then a BRCA1 ChIP is treated as reporting the dimer and nobody sees a reason to ChIP BARD1. On that reading BARD1's occupancy is not missing data but inferred data - which is exactly the inference this graph should not make, since the BRCA1-independent functions of BARD1 are an active literature. Q32-brca1-hr-licensing 2026
Connected to