Oncofusion

start / graph / ana-q125-the-matched-inputs-give-a-first-clonal-drift-screen-and-make-the-cpg-island-test-interpretable

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q125-the-h2ak119ub-arm-of-the-pcgf1-knockout

Four matched input tracks, used for the two things an input is actually for: no chromosome-scale copy-number difference between the CRISPR clone and its control (max 1.0988, min 0.9470 in both processing states), and the knockout's H2AK119ub depletion from CpG islands is REAL because its own genomic DNA is enriched there, not depleted

also called the clonal-drift screen · what six inputs change · input-corrected CpG localisation

A CRISPR-edited cell clone can quietly gain or lose whole chromosomes, which would fake any comparison with its parent. Sequencing the raw DNA of both shows they match. The same raw-DNA tracks also show that a difference seen in the edited cells is genuine rather than an artefact of how the DNA was sequenced.

The same thing, for a scientist

**This corpus has repeatedly caveated occupancy claims for want of an input track. The NTvKO arm deposits six and four are readable** (`GSM8123400`, `GSM8123404`, `GSM8123415`, `GSM8123416`; `GSM8123413` and `GSM8123414` are `.filt.bam` only, with no `.tdf`, and this project has no instrument for them) [E1]. Two of the four were already binned by Q124 and were reused without re-download. **1. The clonal-drift screen (criterion N1), the first this project has run.** Per-chromosome share of each input over `U_m`, ratio knockout/control, computed WITHIN each processing state so a difference appearing only across states would be visible as processing: - **maximum 1.0988 (chr14), minimum 0.9470 (chr15)**; **no chromosome outside 1/1.5-1.5 in either state**, so **no chromosome-scale copy-number difference between the clones** [E1]. - Sub-threshold but reproducible in both states and recorded rather than interpreted: chr13 up (1.031 / 1.083), chr14 up (1.099 / 1.053), chr15 down (0.947 / 0.956), chr19 down (0.959 / 0.968). **2. The control that makes the drug-free knockout test interpretable (criteria V3, V4).** PRC1.1 is targeted to unmethylated CpG islands through KDM2B's CxxC domain, so losing PCGF1 should move H2AK119ub OFF islands - a test that needs no drug and no fusion. In DMSO alone, over the 19,870 bins of `U_m` overlapping a UCSC mm10 CpG island: | | H2AK119ub share, control | knockout | ChIP ratio | **input ratio** | input-corrected | |---|---:|---:|---:|---:|---:| | `paired` | 0.036089 | 0.033850 | 0.9380 | **1.0701** | **0.8765** | | `rep2` | 0.036837 | 0.031719 | 0.8611 | **1.0434** | **0.8253** | **The inputs move the OPPOSITE way.** The knockout clone's own genomic DNA is slightly MORE represented at CpG islands (1.0701 and 1.0434), so the ChIP-level depletion cannot be mappability, sequencing bias or copy number, and correcting for the input makes it **larger**: mean **0.8509** against a pre-registered threshold of 0.80, range 0.0512 [E2]. **The criterion does not fire**, and its pre-registered informativeness clause said its non-firing must never be read as 'the knockout failed'. What it establishes is the direction, consistently in two conventions, against an input that argues the other way.

Go deeper
What an input does NOT buy, stated because the corpus keeps hoping otherwise

An input cannot supply a scale factor, so it does not rescue the missing spike-in and every number here is still a within-sample share, blind to uniform change. Re-scoring every headline ratio against each genotype's own input changes it by at most 0.0228 (criterion N2, matched within processing state), and re-scoring over U_own - the 1,167,555 bins positive in all four readable inputs, 95.10% of U_m - changes it by at most 0.0035 (criterion N3). The input matters here for the two structural questions above and barely at all for the ratios, which is worth knowing before the next beat spends downloads on inputs it does not need.

The two unreadable inputs are not worthless: hyp-q117's PEAK-ARM falsifier names GSM8123413 and GSM8123414 as the matched input for a MACS2 call on re-aligned reads, which is exactly what a BAM is for and exactly what this instrument is not. They are deposited in the right format for the one analysis this project has deferred eight cycles.

The other marks' CpG-island ratios, which behave as the mechanism predicts

In DMSO, knockout over control, at the same CpG-island set: H3K27me3 0.9198 - the same direction as H2AK119ub, consistent with PRC2 recruitment downstream of H2AK119ub via JARID2/AEBP2 - and H3K4me3 1.2638, the opposite way. The knockout's islands carry less of both repressive marks and more of the active one. H3K4me3 and H3K27me3 moving in opposite directions at CpG islands is the signature of a shifted bivalent balance rather than a global chromatin failure, which is the direction Sparbier et al. described for menin inhibition and is here produced by a genotype instead.

Facts and figures
scriptanalyses/q125-the-h2ak119ub-arm-of-the-pcgf1-knockout/run.py
statusrun
result summaryNo chromosome-scale CNV between the clones (max 1.0988, min 0.9470). The knockout's H2AK119ub is depleted from CpG islands by 6-14%% while its input is ENRICHED there by 4-7%%, so the input-corrected depletion is 0.8253-0.8765, mean 0.8509.
absence classmeasured-not-analysed
absence class noteThe four readable inputs of GSE260742's NTvKO arm were public from 2025-11-23 and had never been compared with each other. The clonal-drift question they answer had been carried as an unresolvable hazard since cycle 83 (Q124's RP3) purely because nobody had opened them. Two of the four were already on disk from Q124's run. What remains NEVER-MEASURED is PCGF1 protein or transcript in these cells, and MEASURED-NOT-ANALYSED is a proper CNV call, which needs the two .filt.bam inputs and a caller this project does not have.
absence class assigned2026-09-23 (Q125)
inputs readable
[
  "GSM8123400 (KO, .filt.bam.tdf)",
  "GSM8123404 (control, .filt.bam.tdf)",
  "GSM8123415 (KO, .tdf, binned by Q124)",
  "GSM8123416 (control, .tdf, binned by Q124)"
]
inputs not readable
[
  "GSM8123413 (.filt.bam only, 879.2 MB)",
  "GSM8123414 (.filt.bam only, 846.4 MB)"
]
How this could bite us back
  • Reading a knockout's phenotype from a single clone with no rescue is how a field acquires a wrong mechanism. This screen rules out the coarsest confounder and leaves the others open, so every PCGF1 conclusion in this cycle is conditional on a clone whose editing was never verified in the deposited material.
Still open
  1. Would a real CNV caller on GSM8123413/8123414 and GSM8123400/8123415 find sub-chromosomal drift? The BAMs are deposited and this project has no aligner or caller; it is a cluster job, not a desk job.
  2. Is the chr13/chr14/chr15/chr19 pattern reproducible in the ChIP tracks' own inputs from other arms of the series? If it appears in F3 and F3_NT too it is a library property of this platform and not the clone.
Every source behind this entry (2)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Of the six inputs GEO lists for the GSE260742 NTvKO arm, GSM8123413 and GSM8123414 carry exactly one supplementary file each and it ends .filt.bam with no .tdf; the four readable ones (GSM8123400, GSM8123404, GSM8123415, GSM8123416) give a per-chromosome knockout/control share ratio over U_m whose maximum is 1.0988 (chr14) and minimum 0.9470 (chr15), with no chromosome outside 1/1.5-1.5 in either processing state.

    What this evidence cannot show

    A 2 kb-binned input at whole-chromosome resolution detects chromosome-scale copy-number change and NOT focal amplification, small deletion, loss of heterozygosity, translocation or any point mutation, including an off-target CRISPR cut. 'No chromosome-scale difference' is therefore a much weaker statement than 'the clones are isogenic'. The declared gate of 1/1.5-1.5 corresponds to roughly a whole-chromosome gain or loss in a diploid and would not detect a sub-chromosomal segment. Only two of the four inputs are in each processing state, so 'within state' is a single pair each.

    Evidence that argues against this (1)
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    In DMSO alone, the PCGF1-knockout clone's H2AK119ub share over the 19,870 CpG-island bins of U_m is 0.9380 and 0.8611 times its control's in the two conventions, while the matched inputs' CpG-island share ratio is 1.0701 and 1.0434 in the opposite direction, giving an input-corrected ratio of 0.8765 and 0.8253, mean 0.8509 against a pre-registered threshold of 0.80.

    What this evidence cannot show

    NO SPIKE-IN NORMALISATION IS POSSIBLE. Q117 verified zero hits for the entire spike-in vocabulary across all 419,824 characters of GSE260742's 158 sample records and no scale factor is deposited, so every cross-condition number here is a WITHIN-SAMPLE SHARE, which cancels the unknown per-sample scale factor exactly and IS BLIND TO A UNIFORM CHANGE BY CONSTRUCTION. Nothing here measures any mark's abundance; everything measures its distribution. EXPOSURE ASYMMETRY: DMSO 72 h and SGC0946 5 uM 72 h are matched, VTP-50469 500 nM is 48 h, so every VTP ratio is a LOWER BOUND on the 72-hour effect - identical in both genotypes, so the genotype DIFFERENCE is unaffected. REAGENTS: H2AK119ub is Ubiquityl-Histone H2A (Lys119) (D27C4) XP Rabbit mAb, NEB 8240S, and it CANNOT distinguish PRC1.1 from canonical PRC1 - RING1A/B is the ligase in both and PCGF1 is one of six PCGF paralogs - nor deposition from loss of a deubiquitinase. The H3K27me3 and H3K4me3 reagents are NOT RESOLVED and are recorded unresolved rather than guessed; `antibod` occurs zero times in the 158 GEO records. S_FLAG is defined by DYKDDDDK Tag (D6W5B) Rabbit mAb, NEB 14793S, on a RETROVIRALLY OVER-EXPRESSED tagged MLLAF9 construct, so it is the over-expressed molecule's sites and not an endogenous fusion's. THE KNOCKOUT is one CRISPR guide, one clone C1, no second clone, no rescue and NO deposited PCGF1 protein or transcript measurement. SYSTEM: murine MLL-AF9 (KMT2A::MLLT3) myeloid blood cells, NOT the infant B-lineage KMT2A::AFF1 disease this graph is about. THE CRITERION DOES NOT FIRE at its pre-registered 0.80 threshold and its pre-registered informativeness clause forbids reading that as 'the knockout failed': a within-sample share is blind to uniform change, so V3 sees the knockout only if PCGF1 loss REDISTRIBUTES the mark, and if PRC1.1's contribution at islands is a minority of total H2AK119ub there a real effect can be far under 20%. The CpG-island annotation is UCSC mm10 cpgIslandExt with dataTime 2021-04-08, a computational prediction from sequence composition and not a measurement of methylation in these cells.

    Evidence that argues against this (2)
Connected to