Oncofusion

start / graph / gene-cbx1

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q35-encode-reanalysis

CBX1 / HP1-beta (P83916, 185 aa) - the HP1 paralogue with the fitness cost and no role in this graph's cascade: DepMap mean -0.349 with 16.6% of 1,257 lines below -0.5, four times deeper than HP1-gamma or HP1-alpha, and zero papers place it anywhere near KMT2A

also called HP1beta · HP1-beta · HP1Hsbeta · M31 · modifier 1 · p25beta · heterochromatin protein 1 homolog beta · CBX

There are three near-identical human versions of a protein called HP1, and cancer cells can lose two of them with almost no ill effect but not the third. Nobody knows why, and the one that matters is the one nobody in this field studies.

The same thing, for a scientist

**Named in 20 entries of this graph and never written, for two cycles running.** This closes it. **The asymmetry that is the reason to write it.** Across the same DepMap release, the same metric and near-identical line counts: | paralogue | accession | length | mean gene effect | % of lines below -0.5 | screens | |---|---|---:|---:|---:|---:| | **CBX1 / HP1-beta** | **P83916** | **185** | **-0.349** | **16.6%** | 1,257 | | CBX3 / HP1-gamma | Q13185 | 183 | -0.082 | 1.1% | 1,258 | | CBX5 / HP1-alpha | P45973 | 191 | -0.083 | 0.4% | 1,258 | **CBX1 is about four times deeper on the mean and fifteen to forty times deeper on the tail**, and it sits alongside MEN1 (-0.362) and PALB2 (-0.339) rather than alongside its own paralogues. Three proteins within eight residues of each other, sharing a chromodomain and a chromo shadow domain and the same H3K9me2/3 ligand, and one of them carries essentially all of the family's fitness cost. **Nothing in this graph or, as far as this beat could find, in the literature explains it.** **The DSB literature names the other two.** The CTCF licensing cascade this graph carries runs through **HP1-gamma** (CBX3); the H3K9me3/H4K20me3 53BP1 work names **HP1-beta**; and a third body of work says only "HP1". So CBX1 is present in the damage field and absent from the fusion field, which is the opposite of the pattern for CBX3. **And it is absent from this disease entirely, as a counted negative.** Run live 2026-09-22 with one `[tiab]` tag per term and every `querytranslation` read back: `CBX1[tiab] AND KMT2A[tiab]` = **0**, `CBX1[tiab] AND MLL[tiab]` = **0**, both with non-empty translations and `warninglist.outputmessages = ['No items found.']`. `CBX1[tiab] AND leukemia[tiab]` returns 4 and `HP1beta[tiab] AND leukemia[tiab]` returns 4, none on topic for a KMT2A fusion. **The vocabulary caveat is severe here and is the point of the registry row**: the literature routinely writes "HP1" with no paralogue letter, so a symbol-keyed zero cannot see a paper that measured HP1-beta and called it HP1. **Newly measured here: CBX1 is not at the KMT2A locus either.** ENCODE holds two released CBX1 TF ChIP-seq experiments (K562, HepG2) and this beat read both. Zero called peaks anywhere in the 90 kb gene; the breakpoint cluster core at 0.70x and 0.95x the genome-wide median of same-width tiles; the exon-11 CTCF element at **0.43x and 0.73x the median CTCF anchor of its own cell type**. So the paralogue with the fitness cost is no more present at this locus than the two without it. **Identifier handling, because this family defeats every standard defence.** Resolved **by accession**, never by symbol or name, exactly as `data/symbol-collisions.json` prescribes. Readback: P83916 -> primary gene `CBX1`, entry name `CBX1_HUMAN`, 185 aa. **A refinement to that registry row, measured here:** the row says name readback fails for this family because neither `CBX3_HUMAN` nor "Chromobox protein homolog 3" contains "HP1" or a paralogue letter. True of the entry name and the *recommended* name - **but the alternative-name lists do disambiguate**: CBX1 carries "HP1Hsbeta" and "Heterochromatin protein 1 homolog beta", CBX3 carries "Heterochromatin protein 1 homolog gamma", CBX5 carries "Heterochromatin protein 1 homolog alpha". Reading `proteinDescription.alternativeNames` resolves the paralogue where reading `recommendedName` does not.

What it looks like
Chromobox protein homolog 1185 aa
Chromo domainChromo domain · residues 21–79Chromo shadow domainChromo shadow domain · residues 117–1750100

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.

Facts and figures
gene symbolCBX1
uniprotP83916
uniprot entry nameCBX1_HUMAN
gene primary readbackCBX1
length AA185
protein nameChromobox protein homolog 1
how the accession was resolvedfetched by ACCESSION P83916 from UniProtKB on 2026-09-22 as data/symbol-collisions.json prescribes for this family, and the primary gene name read back as CBX1. A gene_exact:CBX1 query was also run and returned exactly one reviewed human record whose primary is CBX1, so both routes agree here - but the accession route is the one recorded, because it is the one that is safe for CBX3 and CBX5 as well.
identifier trap checkedthe CBX/HP1 family row in data/symbol-collisions.json. Confirmed live: lengths 185 (CBX1) / 183 (CBX3) / 191 (CBX5) - a CBX1/CBX3 swap is two residues; CBX5 carries the gene synonym HP1A and CBX3 carries none. REFINEMENT ADDED BY THIS BEAT: the alternative-name lists DO carry the paralogue letter (HP1Hsbeta, ...homolog beta / gamma / alpha) even though the recommended names and entry names do not, so alternativeNames is a working readback field for this family and recommendedName is not.
domains sourceUniProt P83916 feature table, fetched 2026-09-22: [{'name': 'Chromo 1', 'start': 21, 'end': 79, 'type': 'Domain'}, {'name': 'Chromo 2; shadow subtype', 'start': 117, 'end': 175, 'type': 'Domain'}]
localizationNucleus (UniProt P83916 subcellular location)
subnuclear localizationheterochromatin and H3K9me2/3 domains; in ENCODE K562 and HepG2 the strongest CBX1 signal among the intervals tested here is at promoters, not at the KMT2A element
localization shown for fusionNO - nobody has looked in any fusion. Newly bounded at the KMT2A locus by this beat in K562 and HepG2 (ENCODE), both negative.
shown for fusionNO - nobody has looked in any KMT2A fusion. The DSB literature that names HP1-beta specifically (PMID 30312172) is in non-haematopoietic cells with ionising radiation and does not touch a fusion.
cell typemeasured here in K562 (CML-derived erythroleukaemic) and HepG2 (hepatoblastoma); no measurement in a haematopoietic progenitor or a KMT2A-rearranged cell
cell stateundamaged, unsynchronised cultured cells in every dataset read here
epigenetic readoutreader of H3K9me2/H3K9me3 via its chromodomain; recruits partners through a PxVxL-binding chromo shadow domain
reader of that markH3K9me2 and H3K9me3
absence classnever-measured
relationship to existing entriesgene-cbx3 and gene-cbx5 exist; this is the third paralogue and the one both of them point at as the unexplained outlier. gene-cbx5's open_questions already record that CBX1 was checked at peak level in K562 by Q110 and was negative; this entry adds the HepG2 experiment, the signal-level and anchor-matched readings, the DepMap decomposition and the UniProt resolution.
gene synonyms in uniprot
[
  "CBX"
]
alternative names in uniprot
[
  "HP1Hsbeta",
  "Heterochromatin protein 1 homolog beta",
  "Heterochromatin protein p25",
  "M31",
  "Modifier 1 protein",
  "p25beta"
]
domains
[
  {
    "name": "Chromo domain",
    "start": 21,
    "end": 79
  },
  {
    "name": "Chromo shadow domain",
    "start": 117,
    "end": 175
  }
]
depmap
{
  "source": "Open Targets Platform 26.06 (meta.dataVersion year 26 month 06), fetched 2026-09-22",
  "gene_effect_metric": "Chronos",
  "n_screens_this_gene": 1257,
  "mean_gene_effect": -0.3494,
  "n_lines_below_minus_0_5": 208,
  "fraction_below_minus_0_5": 0.1655,
  "min": -0.9193,
  "max": 0.3382,
  "approvedSymbol_readback": "CBX1",
  "release_note": "rule 83: DepMap ships quarterly and the payload carries no release id, so the Open Targets dataVersion is recorded instead of a bare fetch date. The screen count is GENE-SPECIFIC - 1257 here - and must not be read as a corpus denominator.",
  "no_median_split": "rule 84: no anti-selective-half tally is reported. Half of any gene set sits above the median by construction and this graph's base rate is 45.3%.",
  "live_low_screen_control": "OR4F5 (ENSG00000186092) queried in the same session returns approvedSymbol OR4F5 with only 3 tissue blocks and 5 screens, confirming the endpoint returns tiny payloads without erroring and that a passing symbol readback is not evidence the essentiality payload is populated. A true zero-block control could not be produced: OR2T1 (ENSG00000175393) returned a null target, which is a different failure."
}
depmap paralogue comparison
{
  "CBX1": {
    "mean": -0.3494,
    "frac_below_minus_0_5": 0.1655,
    "n": 1257
  },
  "CBX3": {
    "mean": -0.0817,
    "frac_below_minus_0_5": 0.0111,
    "n": 1258
  },
  "CBX5": {
    "mean": -0.083,
    "frac_below_minus_0_5": 0.004,
    "n": 1258
  },
  "reference_points_same_call": {
    "MEN1": -0.3616,
    "BRCA1": -0.4888
  },
  "note": "all four fetched in one call, one release, so the differences are not release drift. The screen counts differ by one between CBX1 (1,257) and the others (1,258), which is ordinary gene-specific variation, not a mixed release."
}
encode occupancy at kmt2a measured by this beat
{
  "experiments": [
    "ENCSR948QLZ (K562, signal ENCFF058NFI)",
    "ENCSR819WZE (HepG2, signal ENCFF895JUC)"
  ],
  "called_peaks_in_kmt2a_gene": [
    0,
    0
  ],
  "bcr_core_fold_over_genomewide_tile_median": [
    0.945,
    0.7
  ],
  "ctcf_element_fold_over_same_cell_ctcf_anchor_median": [
    0.728,
    0.429
  ],
  "verdict": "NO OCCUPANCY ABOVE BACKGROUND in either cell type",
  "analysis": "ana-q35-brca1-cbx3-and-cbx1-are-not-at-the-kmt2a-breakpoint-cluster-in-any-encode-cell"
}
bounded absence 2026 09 22
{
  "instrument": "NCBI E-utilities esearch over PubMed; one [tiab] tag per term (rule 94); every querytranslation read back; every zero required a non-empty translation plus warninglist.outputmessages = ['No items found.']",
  "CBX1[tiab] AND KMT2A[tiab]": {
    "count": 0,
    "querytranslation": "\"CBX1\"[Title/Abstract] AND \"KMT2A\"[Title/Abstract]"
  },
  "CBX1[tiab] AND MLL[tiab]": {
    "count": 0,
    "querytranslation": "\"CBX1\"[Title/Abstract] AND \"MLL\"[Title/Abstract]"
  },
  "CBX1[tiab] AND leukemia[tiab]": {
    "count": 4,
    "on_topic": 0
  },
  "HP1beta[tiab] AND leukemia[tiab]": {
    "count": 4,
    "on_topic": 0
  },
  "name_list_searched": [
    "CBX1",
    "HP1beta"
  ],
  "names_NOT_searched_and_why_that_matters": "HP1-beta, HP1B, M31, CBX, p25beta and the unqualified 'HP1' were not run. The registry row for this family exists because the literature writes 'HP1' with no paralogue letter, so this zero is bounded by the name list and not by the literature (rule 70/71).",
  "absence_class": "never-measured"
}
Still open
  1. Why does CBX1 carry a fitness cost four times deeper than CBX3 or CBX5 when the three share a chromodomain, a chromo shadow domain and the same H3K9me2/3 ligand? The first thing to exclude is guide cross-reactivity between three highly similar paralogues; the second is simple abundance. Neither has been checked in this graph.
  2. Is the CBX1 DepMap tail (16.6% of lines below -0.5) enriched in any lineage, and specifically in leukaemia lines or KMT2A-rearranged lines? The per-tissue blocks are returned by the same endpoint and were not decomposed here.
  3. Which HP1 paralogue does the unqualified 'HP1' of the DSB literature actually mean in each paper? Until that is resolved paper by paper, every HP1 claim in this graph is a claim about an unspecified member of a three-member family.
  4. Does CBX1 have a PxVxL-mediated contact with any KMT2A fusion partner? MLLT10/AF10 was proposed for HP1 in Drosophila two decades ago and the human PxVxL inspection has never been done.
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    CBX1 is a 185-residue human chromobox protein (HP1-beta) carrying a chromodomain and a chromo shadow domain, with UniProt accession P83916 and primary gene name CBX1, distinguishable from CBX3 (183 aa) and CBX5 (191 aa) only by accession, primary gene name or alternative-name list.

    What this evidence cannot show

    A database record is a curated summary. Domain boundaries are UniProt's calls on the canonical isoform and were not checked against a structure here.

    Evidence that argues against this (1)
    • This family is the worked example of accession verification failing: three paralogues within eight residues, so the length check in verify-accessions.py cannot catch a CBX1/CBX3 swap (two residues), and rule-65 name readback on the recommended name returns nothing that names the paralogue. UniProt Consortium 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    In DepMap CRISPR screens CBX1 has a mean Chronos gene effect of -0.349 across 1,257 lines with 16.55% of lines below -0.5, against -0.082 for CBX3 (1,258 lines, 1.1%) and -0.083 for CBX5 (1,258 lines, 0.4%), all four figures from one call against one release.

    DepMap CRISPR gene-effect (Chronos) for CBX1, served by the Open Targets Platform GraphQL API · Broad Institute DepMap / Open Targets Platform · Open Targets Platform 26.06 (meta.dataVersion year 26 month 06), fetched 2026-09-22 2026
    What this evidence cannot show

    Rule 83: DepMap ships quarterly and returns no release id in the payload, so the Open Targets dataVersion 26.06 is recorded in its place. Screen counts are gene-specific. A CRISPR gene effect measures fitness in cultured cancer lines over ~21 days and is not a statement about a normal haematopoietic progenitor, about an infant, or about what a drug would do. No median-split proportion is reported (rule 84).

    Evidence that argues against this (1)
    • A deeper mean gene effect can reflect an off-target or copy-number artefact rather than biology, and paralogue-specific CRISPR results are particularly exposed to guide cross-reactivity between near-identical sequences. Chronos corrects for copy number but not for paralogue guide promiscuity, and this family is 183-191 aa with high sequence similarity - so the CBX1/CBX3/CBX5 contrast is exactly the shape in which that artefact would appear. Broad Institute DepMap / Open Targets Platform 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    HP1 paralogues bind partner proteins through a PxVxL motif docking onto the chromo shadow domain dimer, which is the structural basis on which any claim that HP1-beta recruits or is recruited by a repair factor must rest.

    PMID 14765118DOI
    What this evidence cannot show

    The structural work is on HP1 chromo shadow domains in general and peptide ligands, not on CBX1 in a DNA-damage context. It establishes the binding mode, not any particular partner in any particular cell.

    Evidence that argues against this (1)
    • A shared docking surface across three paralogues predicts redundancy, which is the opposite of what the DepMap asymmetry shows. Either the paralogues differ in abundance, localisation or a partner set the PxVxL mode does not capture, or the DepMap difference is technical - and this entry cannot currently distinguish those. Broad Institute DepMap / Open Targets Platform 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    HP1-beta is named specifically, rather than as unqualified 'HP1', in the H3K9me3/H4K20me3 chromatin landscape described for 53BP1 recruitment to DNA lesions - so CBX1 is present in the double-strand-break literature while being absent from the KMT2A fusion literature entirely.

    H3K9me3 and H4K20me3 represent the epigenetic landscape for 53BP1 binding to DNA lesions. · Svobodová Kovaříková A, et al. · Aging (Albany NY) 2018
    PMID 30312172DOI
    What this evidence cannot show

    Non-haematopoietic cells, ionising-radiation-induced damage, and an imaging-led readout; no KMT2A, no fusion, and no locus-level occupancy. It is cited here for the fact that this paralogue is named in the DSB field, not for any mechanism at this locus.

    Evidence that argues against this (1)
    • The CTCF licensing cascade this graph carries runs through HP1-GAMMA, not HP1-beta, and a third body of DSB work writes only 'HP1'. So which paralogue the damage field is actually about is unsettled, and the registry row for this family exists precisely because that cannot be resolved from the papers. Broad Institute DepMap / Open Targets Platform 2026
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    Neither of the two released ENCODE CBX1 ChIP-seq experiments (K562, HepG2) shows any called peak in the KMT2A gene, and at signal level the breakpoint cluster core reads 0.70x and 0.95x the genome-wide median of same-width tiles while the exon-11 CTCF element reads 0.43x and 0.73x the median CTCF anchor of the same cell type.

    What this evidence cannot show

    Two experiments, two non-progenitor cell lines, undamaged resting state. The HepG2 experiment is single-replicate. This bounds the resting state in two cell lines and says nothing about the damage state or about a haematopoietic progenitor.

    Evidence that argues against this (1)
    • Q110 reached the same peak-level conclusion for CBX1 in K562 one cycle earlier using a different peak file, so this is a reproduction rather than a new observation for that half; the HepG2 experiment, the signal-level reading and the anchor-matched null are new. ENCODE Project Consortium 2026
Connected to