Oncofusion

start / graph / gene-bard1

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q32-brca1-hr-licensing

BARD1 (Q99728, 777 aa) - BRCA1's obligate partner, and the member of the cascade that actually touches the chromatin: its PxVxL motif at 569-573 binds HP1, and CTCF binds BARD1 rather than BRCA1 in an undamaged cell

also called BARD1 · Q99728 · BARD1_HUMAN · BRCA1-associated RING domain protein 1 · BRCA1-associated RING domain 1

BARD1 is the permanent partner of BRCA1, and it is the half of the pair that grips the silenced chromatin around a DNA break. In an undamaged cell the loop protein CTCF holds onto BARD1, not BRCA1, so BARD1 is where this whole repair cascade is physically anchored.

The same thing, for a scientist

**Identity.** Q99728, `BARD1_HUMAN`, **777 aa**, primary gene name BARD1, **no UniProt gene synonyms at all**, 11 PDB entries. RING-type zinc finger 50-87, a BRCA1-interaction region 26-119, four ankyrin repeats (427-459, 460-492, 493-525, and a degenerate fourth at 526-546), a flexible linker 554-558, and tandem BRCT domains at **560-653** and **667-777** [E1]. `gene_exact:BARD1 AND organism_id:9606 AND reviewed:true` returned **exactly one** reviewed record, `gene_primary` reading back BARD1. **Why BARD1 and not BRCA1 is the anchoring subunit here.** Three specific results from the 2026 cascade paper, all read in full text rather than from the abstract [E2]: 1. **CTCF associates with BARD1 but *rarely* BRCA1 in the absence of DNA damage.** Etoposide at 50 microM for 1 h augments the CTCF-BARD1 interaction and additionally enables a CTCF-BRCA1 interaction, which remains much weaker than the BARD1 one. 2. **The contact is to CTCF's zinc-finger region** (constructs N 1-267, ZF 246-588, C 578-727), and the zinc fingers that bind HP1-gamma are **not** the ones that bind BARD1 directly - so CTCF holds HP1-gamma and BARD1 through separable surfaces of the same region. 3. **BARD1 reaches the silenced chromatin through an HP1 chromoshadow-domain contact.** The HP1-binding **PxVxL motif is at residues 569-573**, overlapping the start of the BRCT repeats; the **PEELI** mutant of that motif largely loses binding to the chromoshadow domain of **all** HP1 subtypes, and in CTCF-depleted cells the PEELI mutant is not rescued and interacts only marginally with CTCF. **I verified the motif position independently and found a numbering discrepancy worth recording.** Fetching Q99728's sequence, residues 569-573 are **PLVLI** - an exact PxVxL match, and the only such motif in the protein at that position, sitting precisely at the start of BRCT1 (560-653). So the motif coordinate checks out against the canonical record. But the same sentence in [E2] gives the BRCT repeats as *'amino acid residues 570-788'*, and **788 exceeds the 777-residue canonical protein**. That span is therefore in some other frame and must not be copied; UniProt's spans are 560-653 and 667-777. **Dependency.** Mean gene effect **-0.5742** across **1,258** screens, **58.2%** of lines below -0.5, 8.5% below -1.0; blood mean **-0.4533** with 42.2% below -0.5 [E3]. **BARD1 is a deeper dependency than BRCA1** in the same call (-0.574 vs -0.489; 58.2% vs 44.0%), which is not the intuition the literature's BRCA1-centric framing gives. **Archive.** ENCODE has **no BARD1 target record at all** - the target.label facet 404s - so unlike BRCA1 there is no public occupancy dataset to interrogate at any locus [E4]. That puts BARD1 in the `never-measured` class for the locus question while BRCA1 is `measured-not-analysed`.

What it looks like
BRCA1-associated RING domain protein 1777 aa
RINGRING-type zinc finger · residues 50–87Interaction with BRCA1Interaction with BRCA1 · residues 26–119ANK ANK 1 · residues 427–459ANK ANK 2 · residues 460–492ANK ANK 3 · residues 493–525ANK 4 (degenerate) · residues 526–546Flexible linker · residues 554–558PxVxL HP1-binding motif (PLVLI) · residues 569–573BRCT 1BRCT 1 · residues 560–653BRCT 2BRCT 2 · residues 667–7770200400600

No three-dimensional model is available for this protein, so the map above is the whole picture. That is itself informative: AlphaFold does not model chains this long, and the parts that have been crystallised are held under the entries for the individual domains.

Go deeper
The PEELI mutant is the reagent this graph should want

Most of the cascade in 2 is established by depletion, and depletion of a chromatin protein changes many things at once. The BARD1 PEELI mutant is different: it is a five-residue substitution at 569-573 that removes one protein-protein contact - BARD1 to the HP1 chromoshadow domain - and leaves the rest of BARD1 intact.

What it shows in 2: PEELI BARD1 is not rescued in CTCF-depleted cells by re-expressing CTCF fragments, and it interacts only marginally with CTCF. So the CTCF-BARD1 association in cells is substantially routed through HP1, not only through the direct zinc-finger contact.

Why it matters here: a separation-of-function allele is what turns 'H3K9 methylation is correlated with BRCA1/BARD1 loading' into 'the HP1 contact is required for it'. If this graph ever tests the licensing hypothesis in a blood cell, PEELI is the specificity control, and it already exists. The caveat that keeps it honest is in 2's own wording: the motif binds the chromoshadow domain of all HP1 subtypes, so PEELI is not a paralogue-selective reagent - it removes HP1-alpha, -beta and -gamma binding together. Paralogue selectivity in this cascade comes from which HP1 CTCF recruits, not from which HP1 BARD1 can bind.

What is NOT established
  • No measurement of BARD1 anywhere in a KMT2A-rearranged cell, in a B-lineage progenitor, or at the KMT2A locus in any cell type.
  • No ENCODE or other public occupancy dataset for BARD1 in any biosample, so the archive route is closed.
  • The CTCF-BARD1 co-immunoprecipitation is from whole-cell extracts of 293T and U2OS; a co-IP is not a demonstration of co-occupancy at a genomic site.
  • BARD1's UniProt localisation comment carries no evidence code, so even the nuclear statement is uncoded in the database this graph joins against.
Facts and figures
uniprotQ99728
uniprot entry nameBARD1_HUMAN
gene symbolBARD1
gene primary readbackBARD1
protein nameBRCA1-associated RING domain protein 1
length AA777
n PDB entries11
how the accession was resolvedgene_exact:BARD1 AND organism_id:9606 AND reviewed:true, fetched live 2026-09-22, returned EXACTLY ONE reviewed human record with gene_primary BARD1 and no gene synonyms. Length asserted at 777 aa. The PxVxL motif position was then checked directly against the fetched sequence rather than taken from the paper - residues 569-573 are PLVLI.
localizationNucleus - and note that BARD1's UniProt subcellular-location comment carries NO evidence code at all, neither ECO:0000269 nor ECO:0000305. That is unusual for a protein with this much literature and it means the database's location statement for BARD1 is weaker than the one for BRCA1, which carries six experimental citations for the nucleus alone.
subnuclear localizationFrom UniProt's uncoded note: 'During S phase of the cell cycle, colocalizes with BRCA1 into discrete subnuclear foci. Can translocate to the cytoplasm. Localizes at sites of DNA damage at double-strand breaks (DSBs); recruitment to DNA damage sites is mediated by the BRCA1-A complex.' [E2] adds a second, BRCA1-A-independent subnuclear route to the same physical site: HP1-gamma-bound, H3K9-methylated chromatin flanking the break, contacted through the PLVLI motif at 569-573. Neither route has been imaged in a haematopoietic cell, and neither has been placed at the KMT2A locus.
localization shown for fusionfalse
cell typeU2OS and HEK-293T only for everything mechanistic in this entry [E2]; the HP1-BARD1 interface was established in the same non-haematopoietic tradition [E5]. 1,258 DepMap lines including 147 haematopoietic and lymphoid [E3]. NO KMT2A-rearranged cell, no infant material, no B-lineage progenitor, and no ENCODE experiment in any biosample [E4].
cell stateS/G2 is written into the UniProt note itself - the BRCA1/BARD1 foci are described as forming during S phase. Combined with the damage-conditional recruitment in [E2], BARD1's relevant occupancy exists only in the intersection of (cycling, S/G2) and (damaged, minutes-scale). The index disease's putative cell of origin is a fetal progenitor whose cycling state at the moment of the translocation is unknown, and that is the single largest gap between this cascade and this disease.
epigenetic readoutREADER, indirectly and through a partner: BARD1 does not bind a histone mark itself in [E2], it binds the chromoshadow domain of HP1-gamma, which binds H3K9me2/me3. So BARD1's chromatin address is one step removed from the mark, and the PEELI mutant is the separation-of-function reagent that proves it. BARD1 also reads poly(ADP-ribose) through its BRCT domain as an independent route to breaks, and reads H4K20me0 as part of the BRCA1/BARD1 heterodimer - neither of which is asserted from a paper read in this beat.
mark read indirectlyH3K9me2/H3K9me3, via the HP1-gamma chromoshadow domain
shown for fusionNO - nobody has looked, in any fusion. BARD1[tiab] AND KMT2A[tiab] returns 0 with a non-empty querytranslation and 'No items found.'; BARD1[tiab] AND 11q23[tiab] returns 0 the same way; BARD1[tiab] AND MLL[tiab] returns 2, of which one is an ASH2L CRISPR screen in glioblastoma and one is an Arabidopsis paper about WUSCHEL1. BARD1[tiab] AND leukemia[tiab] returns 5, none about KMT2A.
absence classnever-measured
absence class reasoningENCODE holds no BARD1 target record at all, so there is no public occupancy dataset that could be re-analysed at the KMT2A locus. Unlike BRCA1 this one cannot be closed with compute; it needs an antibody and an experiment.
domains
Interaction with BRCA126.0Interaction with BRCA1: 26RING-type zinc finger50.0RING-type zinc finger: 50ANK 1427.0ANK 1: 427ANK 2460.0ANK 2: 460ANK 3493.0ANK 3: 493ANK 4 (degenerate)526.0ANK 4 (degenerate): 526Flexible linker554.0Flexible linker: 554BRCT 1560.0BRCT 1: 560PxVxL HP1-binding motif (…569.0PxVxL HP1-binding motif (PLVLI): 569BRCT 2667.0BRCT 2: 667
gene synonyms in uniprot
[]
identifier trap found here
{
  "trap": "BARD1 is also a PLANT gene, and the plant literature collides with this project's core token MLL",
  "measured_2026_09_22": "BARD1[tiab] AND MLL[tiab] returns 2 records, and one of them is PMID 19704708, 'BARD1 may be renamed ROW1 because it functions mainly as a REPRESSOR OF WUSCHEL1' (Plant Signaling & Behavior 2009). Arabidopsis BARD1 is a shoot-meristem regulator with no relation to Q99728, and 'MLL' in that context is not KMT2A.",
  "why_it_matters": "A beat bounding a BARD1 absence on a count of 2 without title-checking would be reporting a literature that is 50% botany. Same family as the SET-NUP214 quantum-chemistry hits and the BENC/BDME case in rule 70.",
  "registry_status": "NOT in data/symbol-collisions.json as of 2026-09-22, checked with the corrected rule-88 command over d['collisions'] (177 rows, case-sensitive - BARD1 returns zero rows). Reported for a row in the hand-back."
}
numbering discrepancy found and resolved
{
  "what_the_paper_says": "[E2] states the HP1-binding PxVxL motif is at 'amino acid residues 569-573' and that it overlaps the start of 'two BRCT repeats (amino acid residues 570-788)'.",
  "what_the_canonical_record_says": "Q99728 is 777 aa. BRCT1 is 560-653 and BRCT2 is 667-777. The span 570-788 runs eleven residues past the end of the protein.",
  "the_check_that_settled_it": "The Q99728 sequence was fetched and scanned for P.V.[LI]. The ONLY match is PLVLI at 569-573 - so the MOTIF coordinate is correct in the canonical frame and the BRCT span quoted alongside it is not. Do not copy 570-788.",
  "rule_this_instances": "the PDB/ASH2L offset family - a residue number that is right in its own frame and wrong in yours. Here the two halves of one sentence are in different frames."
}
domains — the numbers behind the chart
[
  {
    "name": "RING-type zinc finger",
    "start": 50,
    "end": 87
  },
  {
    "name": "Interaction with BRCA1",
    "start": 26,
    "end": 119
  },
  {
    "name": "ANK 1",
    "start": 427,
    "end": 459
  },
  {
    "name": "ANK 2",
    "start": 460,
    "end": 492
  },
  {
    "name": "ANK 3",
    "start": 493,
    "end": 525
  },
  {
    "name": "ANK 4 (degenerate)",
    "start": 526,
    "end": 546
  },
  {
    "name": "Flexible linker",
    "start": 554,
    "end": 558
  },
  {
    "name": "PxVxL HP1-binding motif (PLVLI)",
    "start": 569,
    "end": 573
  },
  {
    "name": "BRCT 1",
    "start": 560,
    "end": 653
  },
  {
    "name": "BRCT 2",
    "start": 667,
    "end": 777
  }
]
depmap
{
  "n_screens_with_gene_effect": 1258,
  "mean_gene_effect": -0.5742,
  "median_gene_effect": -0.5604,
  "min_gene_effect": -2.1418,
  "max_gene_effect": 0.5401,
  "frac_below_minus_0_5": 0.5819,
  "frac_below_minus_1_0": 0.0851,
  "n_blood_screens": 147,
  "blood_tissue_blocks": [
    "lymphoid tissue",
    "hematopoietic tissue"
  ],
  "blood_mean_gene_effect": -0.4533,
  "blood_frac_below_minus_0_5": 0.4218,
  "n_tissue_blocks_flattened": 29,
  "open_targets_data_version": {
    "year": "26",
    "month": "06",
    "iteration": null
  },
  "depmap_release": null,
  "rule_83": "no DepMap release identifier is present in the payload; the Open Targets data version is recorded instead and is not a substitute.",
  "rule_84": "no median-split proportion is reported.",
  "live_zero_screen_control": "POLR2A in the same call returned 0 tissue blocks and 0 screens with a passing approvedSymbol readback, confirming that an empty row would have been visible as empty rather than as a flat dependency.",
  "reading": "BARD1 is DEEPER than BRCA1 in the same call and the same data version - mean -0.574 against -0.489, and 58.2% of lines below -0.5 against 44.0%. The pair is obligate, so the asymmetry is informative rather than contradictory: it is consistent with BARD1 having heterodimer-independent functions, and it means a BARD1-directed perturbation is NOT a milder version of a BRCA1-directed one."
}
How this could bite us back
  • BARD1 is a deeper CRISPR dependency than BRCA1 (-0.574 vs -0.489, 58.2% vs 44.0% of 1,258 lines below -0.5), so any reasoning that treats the heterodimer as one unit with BRCA1's fitness cost understates what removing the BARD1 half would do.
  • The PEELI reagent that makes this cascade testable is NOT paralogue-selective - it removes binding to all three HP1 chromoshadow domains - so an experiment using it cannot attribute a phenotype to HP1-gamma specifically, and this graph must not report a PEELI result as an HP1-gamma result.
  • BARD1 has no ENCODE occupancy data in any cell, so unlike BRCA1 the cheap compute-only route to the locus question does not exist for it; any plan that assumes both halves of the heterodimer can be checked the same way is wrong about cost.
Still open
  1. Is BARD1 ChIP-seq deposited anywhere outside ENCODE? A GEO and ChIP-Atlas sweep was NOT run in this beat and would move this absence from never-measured to measured-not-analysed for the price of one query - exactly the class of named-but-unchecked source the standing brief says to close.
  2. Does the PLVLI motif at 569-573 matter at a TOP2-induced break as well as at a FokI-induced one? Every PEELI experiment in [E2] uses nuclease or laser damage; etoposide is used only for the co-immunoprecipitations.
  3. Why is BARD1 a deeper dependency than its obligate partner? If BARD1 has heterodimer-independent essential functions, then 'BRCA1/BARD1' as a single node in this cascade is an oversimplification the graph is importing from the source paper.
  4. Does CTCF bind BARD1 at the KMT2A exon-11 CTCF element specifically, or only at breaks? The undamaged-state CTCF-BARD1 association reported in [E2] is a whole-cell co-IP with no genomic address, and CTCF's address at this locus is the one thing this graph knows well.
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    BARD1 is UniProt Q99728, entry name BARD1_HUMAN, 777 amino acids, primary gene name BARD1 with NO gene synonyms; the feature table annotates a RING-type zinc finger at 50-87, a BRCA1-interaction region at 26-119, ankyrin repeats ANK1 427-459, ANK2 460-492, ANK3 493-525 and a degenerate ANK4 526-546, a flexible linker 554-558, and BRCT domains at 560-653 and 667-777; it cross-references 11 PDB entries; and its subcellular-location comment carries no evidence code, with an uncoded note stating that BARD1 colocalizes with BRCA1 into discrete subnuclear foci during S phase, can translocate to the cytoplasm, and localizes at DSBs with recruitment mediated by the BRCA1-A complex. Scanning the fetched sequence for P.V.[LI] returns exactly one match, PLVLI at residues 569-573.

    What this evidence cannot show

    A database record fetched 2026-09-22. The motif scan is my own computation over the fetched sequence, not a UniProt annotation - UniProt does not type PLVLI as a feature, so the 569-573 entry in data.domains is sourced from the paper and confirmed by sequence, not from the feature table. The absence of an evidence code on the subcellular-location comment is recorded as observed and is not interpreted as doubt about nuclear localisation.

    Evidence that argues against this (1)
    • UniProt types BRCT1 as beginning at 560 while [E2] describes the BRCT repeats as 570-788. The database and the primary paper disagree about this span, and the paper's upper bound exceeds the protein's length, so at least one of the two is in a non-canonical frame. This entry follows UniProt for the domain spans and the paper only for the motif, which the sequence independently confirms. Hwang SY et al. 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In U2OS and HEK-293T cells, CTCF associates with BARD1 but rarely with BRCA1 in the absence of DNA damage, and treatment with 50 micromolar etoposide for 1 h augments the CTCF-BARD1 interaction and additionally enables a much weaker CTCF-BRCA1 interaction; CTCF binds BARD1 through its zinc-finger region (fragments N 1-267, ZF 246-588, C 578-727) and the zinc fingers that interact with HP1-gamma are unable to bind BARD1 directly; the HP1-binding PxVxL motif of BARD1 lies at residues 569-573 and the PEELI substitution of it largely abolishes binding to the chromoshadow domain of all HP1 subtypes, with the PEELI mutant interacting only marginally with CTCF and failing to be rescued in CTCF-depleted cells; and within the wider cascade CTCF recruits HP1-gamma, KAP1, SUV39H1, SETDB1 and KDM5A to enable rapid BRCA1/BARD1 accumulation at DSBs, whose loss abrogates EXO1/DNA2 recruitment, end resection and HR.

    CTCF primes repressive chromatin for rapid BRCA1/BARD1 recruitment to drive homologous recombination repair of DNA double-strand breaks. · Hwang SY, Choi D, Choi MJ, Jo HR, Jeong JH, Kang MA, Lee JS · Nucleic Acids Research 2026
    PMID 42475683DOI
    What this evidence cannot show

    U2OS osteosarcoma reporter lines and HEK-293T only. The CTCF-BARD1 and CTCF-HP1-gamma contacts are co-immunoprecipitations from whole-cell extracts plus GST pull-downs with bacterially expressed CTCF fragments; a co-IP establishes association, not co-occupancy at a genomic site, and the truncation mapping is done on over-expressed tagged constructs. The BRCT span quoted in the same sentence as the motif (570-788) exceeds the canonical 777-residue protein and should not be reused. DOI taken from PubmedData/ArticleIdList only.

    Evidence that argues against this (1)
    • The PxVxL/HP1 route is not the only published BARD1 recruitment mechanism and may not be the dominant one: the BRCT domain of BARD1 binds poly(ADP-ribose) to recruit BRCA1/BARD1 to DSBs, and RNF168-mediated H2AK15 ubiquitination provides a third route - both described in this paper's own introduction. An HP1-dependent step can be necessary in a reporter system and quantitatively minor in a physiological break. Hwang SY et al. 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    BARD1 is a broad CRISPR dependency and a deeper one than BRCA1 in the same data version: flattening every tissue block gives 1,258 screens with a gene effect, mean -0.5742, median -0.5604, minimum -2.1418, maximum 0.5401, with 58.2% of lines below -0.5 and 8.5% below -1.0; in the 147 hematopoietic and lymphoid screens the mean is -0.4533 with 42.2% below -0.5. BRCA1 in the same call gives mean -0.4888 with 44.0% below -0.5.

    Open Targets Platform GraphQL API, target.depMapEssentiality with all tissue blocks flattened, plus meta.dataVersion · Open Targets Platform; Broad Institute DepMap · Open Targets Platform 2026
    What this evidence cannot show

    Rule 83: the payload carries no DepMap release identifier; the Open Targets data version is year 26 month 06, which is not a release. Rule 84: no median-split proportion is reported. The live zero-screen control fired in the same call - POLR2A returned 0 blocks and 0 screens with a passing symbol readback. BRCA1 and BARD1 have the same screen count here (1,258), so the comparison between them is not confounded by differing denominators; that is not true of DNA2 in the same panel.

    Evidence that argues against this (1)
    • A CRISPR screen in unstressed culture is the condition in which an HR gene is least required, and a knockout is not a model of a transient chromatin-licensing block. The BARD1-deeper-than-BRCA1 asymmetry could also reflect differing guide efficiency or differing tolerance of hypomorphic alleles rather than differing biological requirement. Open Targets Platform; Broad Institute DepMap 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    ENCODE holds no released TF ChIP-seq experiment for BARD1 and no target record for it at all: the target.label facet returns HTTP 404, as it does for EXO1, DNA2, SUV39H1 and RBBP8. In the same census BRCA1 returns 7 experiments (GM12878, H1, HeLa-S3, HepG2, K562), CBX3 4 (GM12878, HCT116, K562), CBX1 2, CBX5 4, KDM5A 3 (A549, H1, HepG2), TRIM28 5, SETDB1 5 and CTCF 502.

    What this evidence cannot show

    Counted 2026-09-22 on the released TF ChIP-seq facet only. A 404 is being read as 'no target record exists', which is what the API's behaviour means, but it is a different kind of negative from a searched zero and is recorded as such. GEO individual depositions, ChIP-Atlas and ReMap were NOT swept, so a BARD1 ChIP-seq could exist outside ENCODE; that is why the absence class for BARD1 is stated as never-measured with respect to ENCODE and the wider archive check is listed as an open question.

    Evidence that argues against this (1)
    • Absence from ENCODE is absence from one consortium, and this project has already been burned by treating one archive route as the archive: the GEO SuperSeries case gave three independent false negatives from three query forms. A GEO sweep on BARD1 ChIP could return depositions, which would move this from never-measured to measured-not-analysed. NCBI Gene Expression Omnibus 2026
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The BARD1-HP1 interface predates the 2026 cascade paper and was established as a requirement for BRCA1 retention at damage sites: interaction of BARD1 and HP1 is required for BRCA1 retention at sites of DNA damage.

    Interaction of BARD1 and HP1 Is Required for BRCA1 Retention at Sites of DNA Damage. · Wu W, Nishikawa H, Fukuda T, et al. · Cancer Research 2015
    PMID 25634209DOI
    What this evidence cannot show

    TITLE-LEVEL ONLY. This reference was read out of the reference list of PMID 42475683 (its ref 21) and the record itself was NOT fetched from PubMed in this beat, so no PMID is asserted, the author list is truncated as printed in that reference list, and nothing is claimed about the paper's system, its HP1 paralogue or its methods. It is cited because [E2] attributes the PxVxL/PEELI reagent and the 'all HP1 subtypes' statement to it, and a reader needs to know the interface is not new. Per the standing rule on identifiers, the DOI here comes from the reference list of another paper and has NOT been verified against PubMed - treat it as unverified until fetched. VERIFIED 2026-09-22 - this citation is no longer unverified; see citation.verification_note. BUT THE PAPER SAYS "HP1" WITH NO PARALOGUE LETTER, in the title and in the abstract sentence that carries the claim. The registry row on the CBX/HP1 family exists for exactly this: CBX1, CBX3 and CBX5 are within 8 residues of one another and the DSB literature names different paralogues in different papers. Which HP1 BARD1 binds is NOT established by this abstract and must not be inferred from the HP1-gamma that PMID 42475683 reports.

    Evidence that argues against this (1)
    • This citation is exactly the shape the brief warns about - a DOI lifted from a reference list, which is how wrong DOI-PMID pairings have entered this graph before. It is recorded with that flag rather than omitted, because the claim it supports (the interface is prior art, not a 2026 discovery) would otherwise be invisible. Any future use must re-fetch it. Hwang SY et al. 2026
Connected to