Oncofusion

start / graph / hyp-cdk13-is-the-lineage-consistent-transcriptional-cdk-dependency

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

The transcriptional CDK with a real KMT2A-rearranged selectivity signal is CDK13, not CDK9 - and nobody has looked

also called CDK13 dependency hypothesis

Screens say KMT2A-rearranged leukaemia cells need CDK13 more than other leukaemia cells do, and unusually the effect is the same in both the lymphoid and myeloid halves of the panel. CDK9, the kinase everyone studies, shows no such pattern. Almost no one has studied CDK13 in leukaemia at all.

The same thing, for a scientist

**Statement.** CDK13, not CDK9, is the transcriptional CDK with a genuine KMT2A-rearrangement-selective requirement, and the CDK9 signal in this disease is floor compression rather than selectivity. **Origin: data pattern.** On this project's genome-scale scan CDK13 ranks 171st of 17,728 (-0.238) and the lineage split gives -0.471 myeloid against -0.492 lymphoid, versus -0.278 in other leukaemia lines and -0.142 across all 1,208 lines. That is the shape of MEN1 (-1.270 / -1.268 vs -0.416) at roughly a third of the amplitude and the opposite of CDK9's shape (-1.826 / -1.555 vs -1.580, 99.3% of all lines depleted). Lineage consistency is the property this project demands after CEBPA and SPI1 failed it, and CDK13 has it. **Prior probability 0.25.** The reasons to be cautious are stated in `gene-cdk13`: ten index lines, a comparator arm already deeper than the all-lines mean, cyclin K shared with CDK12, and CDK13 lying at 7p14.1 within a region deleted in at least one myelomonocytic leukaemia line. **Literature: essentially none.** An alias-aware search returns four leukaemia records and no functional study; a KMT2A-specific search returns one incidental record.

Go deeper
The comparison that generated it
geneKMT2A-r myeloidKMT2A-r lymphoidother leukaemiaall lineshalves agree?scan rank
MEN1-1.270-1.268-0.416-0.363yes1
KMT2A-0.802-0.706-0.293-0.206yes11
CDK13-0.471-0.492-0.278-0.142yes171
CCNT1-0.485-0.085-0.168-0.206no561
CDK9-1.826-1.555-1.580-1.536n/a, floor1,136
CDK12-0.289-0.308-0.390-0.410yes, wrong sign17,136

CCNT1 is in the table to show what failing the split looks like: a large myeloid number and nothing in the lymphoid half. CDK13 passes. So does the reference axis. CDK9 cannot be assessed because 99.3% of all lines are depleted.

Facts and figures
statementCDK13 is the transcriptional CDK with a genuine, lineage-consistent KMT2A-rearrangement-selective requirement; CDK9's apparent depth in this arm is floor compression.
origindata-pattern
prior probability0.25
predictions
[
  "CDK13 knockdown reduces viability of KMT2A::AFF1 B-ALL lines (SEM, RS4;11) more than of matched non-KMT2A B-ALL lines.",
  "CDK13 knockdown reduces KMT2A fusion target-gene output (HOXA9, MEIS1) more than CDK12 knockdown at equal knockdown efficiency.",
  "A CDK13-selective degron or chemical probe separates the effect from CDK9 and from CDK12.",
  "The dependency is preserved in KMT2A-rearranged patient-derived xenografts, where the myeloid/lymphoid confound can be controlled by design."
]
experiments
[
  "exp-cdk13-versus-cdk12-knockdown-lineage-matched-kmt2a-panel"
]
analyses
[
  "ana-ptefb-reservoir-enumeration"
]
falsifiers
[
  "CDK13 knockdown in KMT2A::AFF1 lines is no more depleting than in matched non-KMT2A B-ALL lines at equal knockdown.",
  "The effect of CDK13 loss is abolished by cyclin K overexpression, showing it acts by depriving CDK12 rather than by CDK13's own function.",
  "Re-running the DepMap contrast with a larger and lineage-balanced KMT2A-rearranged panel drops CDK13 out of the top 1,000.",
  "CDK13 knockdown does not reduce HOXA9 or MEIS1 output, which would place it outside the fusion programme whatever it does to viability."
]
How this could bite us back
  • CDK13 is also required in other leukaemia lines (-0.278) and mildly across all lines, so even if the selectivity is real the window is narrow.
  • Cyclin K is shared with CDK12, whose loss suppresses BRCA1 and the Fanconi genes; any inhibitor with cross-reactivity induces a DNA-repair defect in a paediatric patient.
  • There is no selective CDK13 tool compound with published leukaemia data, so this hypothesis currently cannot be tested pharmacologically at all - only genetically.
Still open
  1. Is there a CDK13-selective chemical probe? Every compound in the literature that hits CDK13 also hits CDK9, and usually HDACs as well.
  2. Does CDK13 phosphorylate the Pol II CTD at fusion target genes, and can that be shown by ChIP for CTD Ser2 after CDK13 depletion?
  3. Would a CDK13 dependency be a KMT2A dependency or a HOXA-programme dependency? Testing in NUP98-rearranged lines, which share the HOXA programme without KMT2A, would separate them.
Every source behind this entry (4)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    CDK13 ranks 171st of 17,728 on index-minus-comparator difference in this project's genome-scale DepMap scan, with -0.471 in the six KMT2A-rearranged myeloid lines and -0.492 in the four lymphoid lines against -0.278 in other leukaemia lines and -0.142 across all 1,208 lines; CDK9 in the same query gives -1.826 / -1.555 against -1.580 with 99.3% of all lines depleted.

    What this evidence cannot show

    Ten index lines. The comparator arm is deeper than the all-lines mean, so some of CDK13's difference is leukaemia rather than KMT2A rearrangement. No permutation p was computed for CDK13. Cyclin K is shared with CDK12 and could mediate the effect.

    Evidence that argues against this (1)
    • CDK13 sits at 7p14.1 and has been reported among genes lost in a dic(4;7) deletion in a myelomonocytic leukaemia cell line; a gene a leukaemia can delete is an uncomfortable fit for a gene a leukaemia needs, though a heterozygous deletion is not a knockout. Pan J et al. 2012
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Cyclin K binds both CDK12 and CDK13; the DNA damage response gene requirement is specific to cyclin K/CDK12, so CDK13 has a distinct, less characterised programme.

    The Cyclin K/Cdk12 complex maintains genomic stability via regulation of expression of DNA damage response genes. · Blazek D, Kohoutek J, Bartholomeeusen K, Johansen E, Hulinkova P, Luo Z, Cimermancic P, Ule J, Peterlin BM · Genes Dev 2011
    PMID 22012619DOI
    What this evidence cannot show

    Human cell lines; CDK13's own programme is defined largely by exclusion in that study.

    Evidence that argues against this (1)
    • Later work reports CDK13 cooperating with CDK12 over global Pol II processivity rather than acting on a disjoint gene set, which would make a CDK13 dependency harder to separate from a CDK12 one. Bosken CA et al. 2014
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    An alias-aware PubMed search on 2026-09-20 over CDK13, CDC2L5, CHED, KIAA1791 and CDC2L returns four leukaemia records, none of which studies CDK13 as a dependency or therapeutic target, and a KMT2A-specific search returns one incidental developmental-genetics record.

    What this evidence cannot show

    Bounded by the symbols searched; PubMed dropped KIAA1791 from its query translation. CHED and CDC2L are short and word-like so every hit was title-checked rather than counted, per this project's precision rule.

    Evidence that argues against this (1)
    • Dual CDK9/12/13 inhibitors with leukaemia activity exist and are indexed under CDK9, so the field has touched CDK13 pharmacologically without ever isolating it. HUGO Gene Nomenclature Committee 2026
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    CDK13 is detected in all seven KMT2A-rearranged panel lines the Human Protein Atlas carries, median 19.6 nTPM, so the dependency is not confounded by differential expression within the panel.

    What this evidence cannot show

    Bulk poly-A cell-line RNA. This controls the index arm internally but does not compare expression between the index and comparator arms, which would be the relevant control and was not run.

    Evidence that argues against this (1)
    • The relevant quantity is cyclin K availability, not CDK13 mRNA, and it was not measured. Blazek D et al. 2011
Connected to