DepMap across the whole CTCF-BRCA1 licensing cascade in one call: CTCF is a near-total pan-essential at -1.83, BARD1 is deeper than BRCA1, the HP1 paralogue that matters is flat while the one that does not is four times deeper - and DNA2's row has 80 screens where everything else has 1,258
also called Q32 DepMap panel · dependency across the resection cascade · the DNA2 denominator anomaly
We looked up how much each gene in this repair chain matters for cancer-cell survival, all in one query so the numbers are comparable. The protein at the top of the chain is essential in almost every cell line tested, which means it can only ever be damaged by a drug, never helpfully tuned.
The same thing, for a scientist
**One call, one Open Targets data version (year 26, month 06), ten genes, with a live zero-screen control.** Every figure below is the flattened mean over all tissue blocks returned by `target.depMapEssentiality`, with the gene's own screen count stated beside it as rule 83 requires. | gene | screens | blocks | mean | % < -0.5 | % < -1.0 | blood mean (n=147) | |---|---:|---:|---:|---:|---:|---:| | **CTCF** | 1,258 | 29 | **-1.8260** | **99.76** | **96.90** | -1.8525 | | **BARD1** | 1,258 | 29 | **-0.5742** | 58.19 | 8.51 | -0.4533 | | **BRCA1** | 1,258 | 29 | -0.4888 | 44.04 | 4.69 | -0.2808 | | **CBX1** (HP1-beta) | 1,257 | 29 | -0.3494 | 16.55 | 0.00 | -0.3187 | | **EXO1** | 1,258 | 29 | -0.1273 | 4.29 | 0.16 | -0.0783 | | **KDM5A** | 1,258 | 29 | -0.1185 | 6.44 | 0.00 | -0.1661 | | **CBX5** (HP1-alpha) | 1,258 | 29 | -0.0830 | 0.40 | 0.00 | -0.0804 | | **CBX3** (HP1-gamma) | 1,258 | 29 | -0.0817 | 1.11 | 0.00 | -0.1088 | | **DNA2** | **80** | **17** | -2.7119 | 100.00 | 100.00 | -2.8876 (n=16) | | POLR2A (control) | **0** | **0** | - | - | - | - | **Four readings, in descending order of how load-bearing they are.** 1. **CTCF is a near-total pan-essential**: 99.76% of 1,258 lines below -0.5 and 96.90% below -1.0. **The top node of this cascade cannot be therapeutically modulated for benefit, only damaged.** Every intervention argument in this beat has to enter below CTCF. 2. **BARD1 is deeper than BRCA1** in the same call - -0.574 against -0.489, 58.2% against 44.0% - on identical denominators, so the comparison is clean. The obligate heterodimer is not two interchangeable halves, and 'BRCA1/BARD1' as one node imports an assumption. 3. **The HP1 paralogue that matters here is the one with no fitness cost.** CBX3/HP1-gamma is the paralogue the cascade uses and is flat at -0.082; CBX5/HP1-alpha is equally flat at -0.083; **CBX1/HP1-beta is roughly four times deeper at -0.349 with 16.55% of lines below -0.5** and has no role in this cascade. A pan-HP1 reagent would pay CBX1's toxicity to get CBX3's effect. 4. **DNA2's row is not comparable and this analysis refuses the comparison.** 80 screens in 17 blocks where everything else has 1,258 in 29; and its panel MAXIMUM (-1.55) is below every other gene's MEAN. Rule 83 says most denominator spread is legitimate, and a sixteen-fold gap combined with a non-overlapping distribution is nonetheless a reason to state the number and not rank on it. **The cause was not diagnosed here.** **The instrument was controlled.** POLR2A, fetched in the same call, returned **0 tissue blocks and 0 screens with its `approvedSymbol` reading back as POLR2A** - the failure mode in which an empty payload reads as 'no dependency' rather than 'no data'. Without that control, DNA2's 80-screen row could not be distinguished from a partially-missing one.
| script | no committed script - a single Open Targets GraphQL request batch run 2026-09-22; the query, the gene list, the Ensembl ids and every returned figure are recorded inline so the measurement can be re-derived rather than only re-quoted |
|---|---|
| status | run |
| run date | 2026-09-22 |
| run by | Q32-brca1-hr-licensing |
| ID resolution note | Every returned record's approvedSymbol was read back and matched the requested symbol, including POLR2A's, which is why its zero row is interpretable as 'no data for a correctly resolved gene' rather than as a failed lookup. |
| rule 83 compliance | No DepMap release identifier is present anywhere in the depMapEssentiality payload. What is recorded is the Open Targets data version and it is explicitly NOT a substitute. Every gene's own screen count is stated beside its figures and no corpus-wide denominator is implied. The 1,257 for CBX1 against 1,258 for its paralogues is a one-screen, gene-specific difference of exactly the kind rule 83 describes and is not a release inconsistency. |
| rule 84 compliance | NO median-split proportion appears anywhere in this analysis. Every fraction is against an absolute Chronos threshold (-0.5 or -1.0), and no gene is described as falling in any half of any ranking. No 'k of n of my genes are in the anti-selective half' tally is computed, because half of every gene set is there by construction and this graph's own base rate for that is 45.3%. |
| blood definition | tissue blocks whose tissueName matched any of 'haematopoietic', 'hematopoietic', 'lymphoid', 'blood' or 'bone marrow' case-insensitively. For all nine populated genes this resolved to exactly two blocks - 'hematopoietic tissue' and 'lymphoid tissue' - totalling 147 screens (16 for DNA2). The matching rule is stated because a blood subset defined by a different string set would give a different n. |
| result summary | CTCF mean -1.826 with 99.76% of 1,258 lines below -0.5 - a near-total pan-essential and therefore untouchable as a therapeutic node. BARD1 (-0.574) deeper than BRCA1 (-0.489) on identical denominators. CBX3 and CBX5 flat (-0.082, -0.083) while CBX1 is ~4x deeper (-0.349) and is the paralogue with no role in the cascade. DNA2's 80-screen row is flagged and excluded from comparison. Live POLR2A zero-screen control fired. |
inputs
[
"Open Targets Platform GraphQL, target(ensemblId).depMapEssentiality{tissueName screens{depmapId cellLineName diseaseFromSource geneEffect expression}}",
"meta{dataVersion{year month iteration}} in the same session"
]ensembl IDS used
{
"BRCA1": "ENSG00000012048",
"BARD1": "ENSG00000138376",
"CBX3": "ENSG00000122565",
"CBX1": "ENSG00000108468",
"CBX5": "ENSG00000094916",
"KDM5A": "ENSG00000073614",
"EXO1": "ENSG00000174371",
"DNA2": "ENSG00000138346",
"CTCF": "ENSG00000102974",
"POLR2A": "ENSG00000181222"
}open targets data version
{
"year": "26",
"month": "06",
"iteration": null
}live zero screen control
{
"gene": "POLR2A",
"ensembl": "ENSG00000181222",
"approvedSymbol_readback": "POLR2A",
"n_tissue_blocks": 0,
"n_screens": 0,
"why_it_matters": "A gene can return zero tissue blocks while its approvedSymbol readback passes, so a genuinely empty row is indistinguishable from a flat dependency to any code that averages an empty list to zero or skips it. Q110 ran the identical control and got the identical result, so this is a reproducible property of the endpoint and not a transient. It is the reason DNA2's small row can be reported as small rather than as suspect-empty."
}results — the numbers behind the chart
{
"CTCF": {
"screens": 1258,
"blocks": 29,
"mean": -1.826,
"median": -1.8235,
"min": -3.0556,
"max": -0.3833,
"frac_below_0_5": 0.9976,
"frac_below_1_0": 0.969,
"blood_n": 147,
"blood_mean": -1.8525,
"blood_frac_below_0_5": 0.9932
},
"BARD1": {
"screens": 1258,
"blocks": 29,
"mean": -0.5742,
"median": -0.5604,
"min": -2.1418,
"max": 0.5401,
"frac_below_0_5": 0.5819,
"frac_below_1_0": 0.0851,
"blood_n": 147,
"blood_mean": -0.4533,
"blood_frac_below_0_5": 0.4218
},
"BRCA1": {
"screens": 1258,
"blocks": 29,
"mean": -0.4888,
"median": -0.4622,
"min": -1.8413,
"max": 0.3641,
"frac_below_0_5": 0.4404,
"frac_below_1_0": 0.0469,
"blood_n": 147,
"blood_mean": -0.2808,
"blood_frac_below_0_5": 0.1769
},
"CBX1": {
"screens": 1257,
"blocks": 29,
"mean": -0.3494,
"median": -0.3467,
"min": -0.9193,
"max": 0.3382,
"frac_below_0_5": 0.1655,
"frac_below_1_0": 0,
"blood_n": 147,
"blood_mean": -0.3187,
"blood_frac_below_0_5": 0.1293
},
"EXO1": {
"screens": 1258,
"blocks": 29,
"mean": -0.1273,
"median": -0.1073,
"min": -1.1343,
"max": 0.3749,
"frac_below_0_5": 0.0429,
"frac_below_1_0": 0.0016,
"blood_n": 147,
"blood_mean": -0.0783,
"blood_frac_below_0_5": 0.0272
},
"KDM5A": {
"screens": 1258,
"blocks": 29,
"mean": -0.1185,
"median": -0.1195,
"min": -0.9972,
"max": 0.7462,
"frac_below_0_5": 0.0644,
"frac_below_1_0": 0,
"blood_n": 147,
"blood_mean": -0.1661,
"blood_frac_below_0_5": 0.1156
},
"CBX5": {
"screens": 1258,
"blocks": 29,
"mean": -0.083,
"median": -0.081,
"min": -0.6567,
"max": 0.5714,
"frac_below_0_5": 0.004,
"frac_below_1_0": 0,
"blood_n": 147,
"blood_mean": -0.0804,
"blood_frac_below_0_5": 0.0068
},
"CBX3": {
"screens": 1258,
"blocks": 29,
"mean": -0.0817,
"median": -0.0749,
"min": -0.6843,
"max": 0.6011,
"frac_below_0_5": 0.0111,
"frac_below_1_0": 0,
"blood_n": 147,
"blood_mean": -0.1088,
"blood_frac_below_0_5": 0.0204
},
"DNA2": {
"screens": 80,
"blocks": 17,
"mean": -2.7119,
"median": -2.7575,
"min": -3.9526,
"max": -1.5485,
"frac_below_0_5": 1,
"frac_below_1_0": 1,
"blood_n": 16,
"blood_mean": -2.8876,
"blood_frac_below_0_5": 1,
"DO_NOT_COMPARE": "80 screens against 1,258; panel maximum below every other gene's mean; cause not diagnosed"
},
"POLR2A": {
"screens": 0,
"blocks": 0,
"note": "live zero-screen control, approvedSymbol readback passed"
}
}what this does NOT show
[ "Selectivity of any kind. No KMT2A-rearranged arm was constructed and no comparator arm was defined, so nothing here is a selectivity claim and no z-score or index-versus-comparator difference is computed. This graph's own history on that (lineage confounding, z-as-narrowness, floor compression) is why the analysis was deliberately kept to absolute gene effects.", "Anything about repair under stress. Every DepMap screen runs without an exogenous genotoxic agent, which is the condition under which an HR gene is LEAST required. A flat score for BRCA1 in blood lines says nothing about what BRCA1 loss costs a progenitor during etoposide exposure.", "Anything about redundancy. CBX1/CBX3/CBX5 are three paralogues and EXO1/DNA2 are a redundant pair; a single-gene knockout score systematically understates pathway importance for both sets, and floor compression does not apply because none of them is near the floor.", "Anything about pharmacology. A CRISPR knockout is total and permanent over a two-week screen; a chemical probe is partial and reversible." ]
- CTCF's 99.76% pan-essentiality forecloses the most obvious therapeutic reading of this whole cascade. Anyone proposing to modulate the top node to protect the KMT2A locus is proposing to damage a gene that 1,255 of 1,258 cell lines cannot lose.
- DNA2's -2.71 is the number most likely to be quoted out of this analysis and the one it least supports. It is recorded with a DO_NOT_COMPARE flag in the data block for exactly that reason, and if it reaches a ledger header it will go stale invisibly - which is the failure this project has already had three times.
- Why does DNA2 return 80 screens where every other gene in the same call returns 1,258? A direct query against the DepMap portal would settle it in minutes and would decide whether a striking number is a finding or an artefact.
- What do these genes look like under genotoxic stress rather than in standard culture? A drug-anchored screen (etoposide or a PARP inhibitor) is the condition in which an HR-licensing panel would actually be informative, and no such arm exists in the data this analysis used.
- Is CBX1's four-fold-deeper dependency explained by something other than redundancy structure? It is the paralogue with no role in this cascade and the only one of the three with a measurable fitness cost, and nothing in this beat explains that.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
Fetched in one Open Targets GraphQL request batch on 2026-09-22 at data version year 26 month 06, with all tissue blocks flattened: CTCF 1,258 screens across 29 blocks, mean gene effect -1.8260, median -1.8235, min -3.0556, max -0.3833, 99.76% below -0.5 and 96.90% below -1.0, blood mean -1.8525 over 147 screens; BARD1 1,258 screens, mean -0.5742, 58.19% below -0.5, blood mean -0.4533; BRCA1 1,258 screens, mean -0.4888, 44.04% below -0.5, blood mean -0.2808; CBX1 1,257 screens, mean -0.3494, 16.55% below -0.5; EXO1 1,258 screens, mean -0.1273; KDM5A 1,258 screens, mean -0.1185, blood mean -0.1661; CBX5 1,258 screens, mean -0.0830; CBX3 1,258 screens, mean -0.0817, blood mean -0.1088; and DNA2 only 80 screens across 17 blocks, mean -2.7119, maximum -1.5485, 100% below -1.0. The live zero-screen control POLR2A returned 0 tissue blocks and 0 screens in the same call with its approvedSymbol reading back as POLR2A.
Open Targets Platform GraphQL API, target.depMapEssentiality with all tissue blocks flattened, and meta.dataVersion, ten genes in one request batch · Open Targets Platform; Broad Institute DepMap · Open Targets Platform 2026What this evidence cannot show
Rule 83: the payload carries NO DepMap release identifier; the Open Targets data version is recorded instead and is not a substitute, so this table cannot distinguish legitimate gene-specific variation from release drift for anyone re-running it later. Rule 84: no median-split proportion is computed anywhere. The blood subset is defined by a stated string match on tissueName and a different string set would give a different n. DNA2's figures are recorded but explicitly excluded from every comparison because of the sixteen-fold denominator gap. All screens are run without exogenous genotoxic stress, which is the least informative condition for a DNA-repair gene panel.
Evidence that argues against this (1)
- Three of the nine populated genes belong to redundant sets - CBX1/CBX3/CBX5 as paralogues and EXO1/DNA2 as a redundant nuclease pair - so their single-gene scores systematically understate requirement, and none of them is floor-compressed, which removes the usual excuse for ignoring a shallow score. The panel therefore reports the fitness cost of losing each gene alone, which is not the quantity any argument in this beat actually needs. Open Targets Platform; Broad Institute DepMap 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The biology that makes the CTCF figure load-bearing: CTCF sits at the top of the licensing cascade, recruiting HP1-gamma, KAP1, SUV39H1, SETDB1 and KDM5A to double-strand breaks to license BRCA1/BARD1 and, downstream, EXO1/DNA2-mediated long-range resection - and the H3K9 methyltransferase inhibitors UNC-0638, chaetocin and Bix-01294 disturb the immediate accumulation of H3K9me3, HP1-gamma and BRCA1/BARD1 at DSBs but NOT of CTCF itself.
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 2026What this evidence cannot show
U2OS and HEK-293T only; no haematopoietic cell, no fusion, no KMT2A locus. Cited here only to establish which node of the cascade the CTCF dependency figure applies to; the inhibitor result is in Supplementary Fig. S5E-G. DOI taken from PubmedData/ArticleIdList only.
Evidence that argues against this (1)
- Pan-essentiality in DepMap is a property of transformed, rapidly cycling cell lines in culture and does not establish that CTCF is equally indispensable in a quiescent normal progenitor. The 'untouchable node' conclusion is well supported for a therapeutic window argument in cancer cells and is weaker as a statement about the cell of origin. Open Targets Platform; Broad Institute DepMap 2026
CTCF mean -1.826, 99.76% of 1,258 lines below -0.5 - the top node is untouchable
BARD1 -0.574 is deeper than BRCA1 -0.489 on identical denominators
80 screens against 1,258 for every other gene in the same call - flagged DO_NOT_COMPARE
CBX3 flat at -0.082 while the uninvolved paralogue CBX1 is four times deeper at -0.349
supplies the fitness-cost layer for every node of the cascade