Oncofusion

start / graph / hyp-the-hp1-paralogue-that-costs-fitness-is-the-one-this-field-never-measures

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q35-encode-reanalysis

The HP1 paralogue asymmetry is a guide-cross-reactivity artefact until somebody shows it is not - and if it is real, HP1-beta is doing a job the other two cannot, which is the only HP1 question in this graph with a fitness phenotype attached

also called the CBX1 asymmetry · why HP1-beta and not HP1-gamma

Three nearly identical proteins, and cancer cells only really need one of them. Either that is a genuine division of labour nobody has explained, or the experiment that measured it was confused by how similar the three genes are - and the cheap test tells you which.

The same thing, for a scientist

**The observation.** CBX1/HP1-beta has a DepMap mean gene effect of **-0.349** with **16.6%** of 1,257 lines below -0.5. CBX3/HP1-gamma is **-0.082** (1.1% of 1,258) and CBX5/HP1-alpha is **-0.083** (0.4% of 1,258). One call, one release, near-identical denominators. Three proteins of 185, 183 and 191 residues sharing a chromodomain, a chromo shadow domain and the same H3K9me2/3 ligand, and one carries essentially all the family's fitness cost. **This entry's primary claim is deliberately deflationary, because the alternative explanation is cheap and has not been excluded.** CRISPR guide libraries targeting three highly similar paralogues are the textbook setting for cross-reactivity, and Chronos corrects for copy number, not for guides that cut more than one family member. **A single-guide-level decomposition of the CBX1 screen - do all guides agree, and do any of them have predicted off-targets in CBX3 or CBX5 - would settle it and has not been done in this graph.** Publishing the asymmetry as biology before running that check would be the same error this project has logged repeatedly: a number that feels like an observation and is a property of the instrument. **If it survives that check, the biological reading is specific and testable.** HP1-beta would be doing something the other two cannot, and the candidates are bounded: differential abundance; differential localisation, with HP1-beta at constitutive heterochromatin where HP1-gamma is euchromatic; or a PxVxL partner that only HP1-beta engages. **The interest for this graph is that the DSB literature names HP1-beta (PMID 30312172) while the CTCF licensing cascade this graph carries names HP1-gamma - so the paralogue with the fitness cost is the one outside the cascade.** If the cascade is right and HP1-gamma is the one that matters at breaks, the fitness cost of the other paralogue is unexplained by everything in this graph. **What makes this more than a restatement.** `gene-cbx3` and `gene-cbx5` both already note that CBX1 is the outlier; neither states what would distinguish an artefact from a division of labour, and neither proposes the guide-level check. The registry row for this family records the numbers and prescribes resolution by accession; it does not ask whether the numbers are real.

Facts and figures
statementThe CBX1-versus-CBX3/CBX5 DepMap asymmetry is more likely to be a CRISPR guide cross-reactivity artefact between three highly similar paralogues than a genuine division of labour, and should not be quoted as biology until a guide-level decomposition of the CBX1 screen has been run.
origindata-pattern
prior probability0.55
prior reasoning0.55 for the artefact reading rather than a confident number in either direction. Against the artefact: 16.6% of 1,257 lines is a large and consistent effect, and guide cross-reactivity usually makes paralogues look MORE similar, not less - a guide that cuts CBX1 and CBX3 would deepen CBX3 too, and CBX3 is flat. For the artefact: the reverse case also exists, where a multi-cutting guide produces a copy-number-like fitness cost attributed to whichever gene it was labelled with, and three genes within eight residues is the worst case for it. The two arguments are close and 0.5-0.6 is the honest range.
what it would change if trueIf the asymmetry is an artefact, this graph should stop treating CBX1 as the interesting paralogue and the registry row for this family should carry the correction, because the -0.349 figure is currently quoted in the registry itself as evidence that the paralogues 'are not interchangeable'. If it is real, then the HP1 member with a fitness phenotype is the one outside this graph's DSB cascade, and the cascade's choice of HP1-gamma needs an argument it does not currently have.
cell typeDepMap cancer cell lines, all lineages; the KMT2A-locus arm is measured in K562 and HepG2 only
cell stateproliferating cultured cells over a ~21-day CRISPR screen window; undamaged for the occupancy arm
epigenetic readoutH3K9me2/H3K9me3 reading by the chromodomain is the shared function whose non-redundancy is in question
shown for fusionNO - nobody has looked at any HP1 paralogue's fitness contribution in a KMT2A-rearranged line specifically; the DepMap tail was not lineage-decomposed in this beat
predictions
[
  "If the asymmetry is real, individual CBX1 guides will agree with each other and none will have a predicted cut site in CBX3 or CBX5.",
  "If the asymmetry is real, CBX1 protein abundance or chromatin occupancy will differ from CBX3 and CBX5 by substantially more than their sequence similarity would suggest.",
  "If the asymmetry is an artefact, the CBX1 effect will be carried by a subset of guides, and those guides will be the ones with family-internal off-targets.",
  "Under either reading, CBX1 will remain absent from the KMT2A locus, which this beat has already measured in K562 and HepG2 - so the asymmetry is not about this locus."
]
experiments
[
  "exp-guide-level-decomposition-of-the-cbx1-depmap-screen"
]
analyses
[
  "ana-q35-brca1-cbx3-and-cbx1-are-not-at-the-kmt2a-breakpoint-cluster-in-any-encode-cell"
]
falsifiers
[
  "FALSIFIER 1 - THE GUIDES AGREE AND ARE CLEAN, SO IT IS NOT AN ARTEFACT. Data: the DepMap public CRISPR screen, per-guide log-fold-change for every CBX1 guide, across the same 1,257 lines. Cell type: all lineages pooled, then leukaemia lines separately. SCORING: if at least 80% of CBX1 guides individually produce a median gene effect below -0.25, AND no CBX1 guide has a perfect or one-mismatch alignment anywhere in CBX3 or CBX5, then the artefact reading is dead and the asymmetry is biological. This hypothesis is then REFUTED in its stated form and the entry must be rewritten around the division-of-labour reading.",
  "FALSIFIER 2 - THE EFFECT IS CARRIED BY CROSS-CUTTING GUIDES. Same data. SCORING: if the CBX1 mean gene effect computed after DROPPING every guide with a family-internal off-target rises above -0.15 - i.e. into the CBX3/CBX5 range of -0.08 - then the asymmetry is an artefact and this hypothesis is CONFIRMED. A rise to between -0.15 and -0.25 is recorded as partial and unresolved.",
  "FALSIFIER 3 - ABUNDANCE EXPLAINS IT AND NO DIVISION OF LABOUR IS NEEDED. System: any cell line in DepMap with matched proteomics, leukaemia lines preferred. SCORING: if CBX1 protein abundance exceeds the sum of CBX3 and CBX5 by more than three-fold across at least ten lines, the asymmetry is a dosage effect rather than a functional one, and the interesting version of this question dissolves.",
  "FALSIFIER 4 - IT IS REAL AND IT IS ABOUT THE DSB CASCADE, WHICH WOULD MATTER HERE. Interval: the KMT2A exon-11 CTCF element, GRCh38 chr11:118,488,159-118,489,206. Cell type: a B-lineage progenitor or CD34+ progenitor, after an induced break within 2 kb. Effect size: if HP1-beta reaches 1.5-fold over a matched chr11 control 20 kb distal at any timepoint while HP1-gamma does not, then this graph's cascade names the wrong paralogue. NOT YET RUN. What HAS been run, in the resting state only: CBX1 is at 0.43x and 0.73x the median CTCF anchor of its own cell type in ENCODE K562 and HepG2, so no resting-state pre-positioning exists - which neither fires nor clears this falsifier, because it specifies the damage state."
]
Still open
  1. Does the DepMap public per-guide table resolve this in one afternoon? It is downloadable and this beat did not fetch it - the Open Targets endpoint this project uses does not expose guide-level data at all, which is itself worth knowing.
  2. Is the CBX1 dependency tail enriched in leukaemia lines? The per-tissue blocks come back from the same query and were not decomposed.
  3. If the asymmetry is real, is it the chromodomain or the chromo shadow domain that carries it? A domain-swap between CBX1 and CBX3 in a rescue assay is the clean version and nobody has run it.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    CBX1 has a mean DepMap Chronos gene effect of -0.349 with 16.55% of 1,257 lines below -0.5, against -0.082 for CBX3 (1.1% of 1,258) and -0.083 for CBX5 (0.4% of 1,258), all fetched in one call against Open Targets dataVersion 26.06.

    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: no DepMap release id in the payload, so the Open Targets dataVersion is recorded; screen counts are gene-specific (1,257 for CBX1, 1,258 for the others). Rule 84: no median-split proportion is reported. A live low-screen control (OR4F5, 5 screens) confirms the endpoint reports sparse payloads as sparse.

    Evidence that argues against this (1)
    • This is exactly the comparison a paralogue guide-cross-reactivity artefact would produce, and nothing in the Open Targets payload reports guide-level data, so the artefact cannot be excluded from this source at all. That is the reason this hypothesis leads with the deflationary reading. Broad Institute DepMap / Open Targets Platform 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    CBX1, CBX3 and CBX5 are 185, 183 and 191 residues, each carrying a chromodomain and a chromo shadow domain, which is the sequence and architectural similarity that makes both the redundancy expectation and the guide-cross-reactivity concern concrete.

    What this evidence cannot show

    Length and domain architecture are not the same as nucleotide-level guide-target similarity, which is what would actually drive cross-cutting and which was not computed here.

    Evidence that argues against this (1)
    • Protein-level similarity can coexist with substantial nucleotide divergence, so the architectural argument for cross-reactivity is suggestive rather than quantitative. It is offered as the reason to run the check, not as evidence that the artefact is present. UniProt Consortium 2026
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The DSB literature names HP1-beta specifically in the H3K9me3/H4K20me3 landscape for 53BP1 recruitment, while this graph's CTCF licensing cascade runs through HP1-gamma - so the paralogue carrying the fitness cost is the one outside the cascade the graph carries.

    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, imaging-led; no KMT2A and no fusion. Cited for which paralogue the damage field names, not for a mechanism.

    Evidence that argues against this (1)
    • A third body of DSB work writes only 'HP1' with no paralogue letter, so the claim that the field names HP1-beta is true of one paper and not of the field. The registry row for this family exists because that ambiguity is unresolvable from the papers. UniProt Consortium 2026
Connected to