Six countermeasures, no head-to-head — so rank their targets: CDK6 is 13th and CDK4 is 16,052nd
also called countermeasure target ranking · resistance combination triage
Six different research groups have each proposed a different drug to add when menin inhibitors stop working, and nobody has compared them. We could not run that comparison, but we could ask a cheaper question of each one: is the protein it targets actually more needed by this leukaemia than by others? Four of the six clear that bar and four do not — and one widely used drug class turns out to contain one target near the top of the list and one near the bottom.
The same thing, for a scientist
The resistance beat now names six countermeasures from six groups with no head-to-head comparison. Looking each target up in this project's persisted 17,728-gene selectivity scan: CDK6 ranks 13, MCL1 18, BCL2 36 and KAT7 92 — all inside the top 1%. BRD4 (3,122), JNK/MAPK8 (4,220), SMARCA2 (11,581) and CREBBP (16,950) do not clear it, and SMARCA4 (16,333), CREBBP and CDK4 (16,052) sit on the *less-required* side. Three of the four that fail are proposed specifically for the resistant state, which DepMap cannot model, so their failure is weak evidence. JNK is not, and its failure is the informative one.
The question this can and cannot answer
A real comparison of six countermeasures needs the experiment that does not exist — exp-orthogonality-time-to-resistance-panel is already written and is what would settle it. But one axis was already paid for: an unbiased scan of 17,728 genes for selective requirement in KMT2A-rearranged versus other leukaemia lines, with every rank persisted.
So the cheap question is available: is this target selectively required in this disease at all? That is a necessary condition, not a sufficient one, and it is screening rather than comparison. The answer is unusually clean — the six do not sit in a band, they split at opposite ends of the list.
The ranking
Rank 1 = most selectively required. Top 1% is rank ≤ 177. A positive difference means the rearranged lines need it less.
| countermeasure | best target | rank of 17,728 | clears top 1%? | proposed for the resistant state only? |
|---|---|---|---|---|
| CDK4/6 | CDK6 | 13 | yes | yes |
| MCL-1 (MIK665) | MCL1 | 18 | yes | no |
| venetoclax | BCL2 | 36 | yes | no |
| KAT6/KAT7 | KAT7 | 92 | yes | no |
| BET | BRD4 | 3,122 | no | yes |
| JNK | MAPK8 | 4,220 | no | no |
| SWI-SNF | SMARCA2 | 11,581 | no | yes |
| CREBBP | CREBBP | 16,950 | no | yes |
For scale, in the same scan MEN1 ranks 1, KMT2A 11, MEIS1 84, AFF1 160 and DOT1L 184.
CDK4 and CDK6 are at opposite ends of the same drug class
| gene | rank | percentile | difference |
|---|---|---|---|
| CDK6 | 13 | 0.073 | -0.5266 |
| CDK4 | 16,052 | 90.546 | 0.0716 |
CDK6 is the 13th most selectively required gene in the genome for this disease. CDK4 is 16,052nd, on the wrong side. They are routinely named together, and every approved inhibitor of the class hits both.
This does not argue against using a CDK4/6 inhibitor — a drug that hits CDK6 hits CDK6. It argues that the rationale should be written as CDK6, that a CDK6-selective agent would be the cleaner instrument if one is available, and that the CDK4 component is carrying toxicity for a target this disease does not preferentially need. An entry recommending 'CDK4/6' is averaging two genes at opposite ends of 17,728.
The one failure that is actually informative
Three of the four countermeasures that fail — BET, SWI-SNF, CREBBP — are proposed specifically for the resistant state, and DepMap profiles unselected KMT2A-rearranged lines. A dependency that only emerges after menin-inhibitor exposure is invisible here by construction. Their low ranks are weak evidence, not refutation, and this analysis is asking them the wrong question.
JNK is not in that category, and it fails anyway. MAPK8 ranks 4,220, MAPK9 11,126, MAPK10 8,079 — nothing.
That matters more than the rest of the table, because JNK inhibition is the proposed countermeasure for JUND liberation, which this graph now holds as the most structurally troubling resistance mechanism it has: the mechanism of action and the mechanism of escape are the same physical event. The most important resistance mechanism currently has the weakest countermeasure support on this axis. That is not a refutation — JNK is a signalling kinase whose relevance is conditional on JUND being liberated, which does not happen in an untreated line — but it is the gap worth closing first.
| script | analyses/countermeasure-target-ranking/run.py |
|---|---|
| status | executed |
| run date | 2026-09-19 |
| n scanned | 17728 |
| top 1pct rank threshold | 177 |
| headline | CDK6 13, MCL1 18, BCL2 36, KAT7 92 clear the top 1%. BRD4, JNK, SWI-SNF and CREBBP do not. CDK4 is 16,052nd while CDK6 is 13th. |
| informative negative | JNK (MAPK8 4,220) is the only failure not explained by the resistant-state design limitation, and it is the countermeasure for JUND liberation - the most structurally troubling resistance mechanism in the graph. |
| caveat | DepMap profiles UNSELECTED KMT2A-rearranged lines. Countermeasures proposed for the RESISTANT state - BET/SWI-SNF for MEN1-wild-type resistant cells, CDK4/6 after KMT2C/KMT2D loss - are being asked the wrong question here, and a low rank for those is weak evidence rather than refutation. Gene effect is not drug effect. Eleven rearranged lines, none infant-derived. A difference statistic is also least informative for strongly pan-essential genes, where a floor effect can compress or invert the sign. |
ranking — the numbers behind the chart
{
"CDK4/6": {
"best_gene": "CDK6",
"rank": 13,
"clears_top_1pct": true,
"resistant_state_only": true
},
"MCL-1 (MIK665)": {
"best_gene": "MCL1",
"rank": 18,
"clears_top_1pct": true,
"resistant_state_only": false
},
"venetoclax": {
"best_gene": "BCL2",
"rank": 36,
"clears_top_1pct": true,
"resistant_state_only": false
},
"KAT6/KAT7": {
"best_gene": "KAT7",
"rank": 92,
"clears_top_1pct": true,
"resistant_state_only": false
},
"BET": {
"best_gene": "BRD4",
"rank": 3122,
"clears_top_1pct": false,
"resistant_state_only": true
},
"JNK": {
"best_gene": "MAPK8",
"rank": 4220,
"clears_top_1pct": false,
"resistant_state_only": false
},
"SWI-SNF": {
"best_gene": "SMARCA2",
"rank": 11581,
"clears_top_1pct": false,
"resistant_state_only": true
},
"CREBBP": {
"best_gene": "CREBBP",
"rank": 16950,
"clears_top_1pct": false,
"resistant_state_only": true
}
}cdk4 vs cdk6
{
"CDK6_rank": 13,
"CDK4_rank": 16052,
"note": "same drug class, opposite ends of 17,728"
}- Ranking a combination partner by single-gene CRISPR selectivity rewards targets that are lethal alone, which is close to the opposite of what a combination is for. A good partner may be one whose loss does nothing by itself and everything alongside menin inhibition - and this analysis would rank it near the bottom.
- Three of the four failures are here because the screen is in the wrong population, and saying so does not undo the impression the table leaves. A reader who takes the ranking at face value would drop the SWI-SNF/BET line of work, which is supported by direct experiments in exactly the cells that matter.
- CDK4/6 inhibitors carry real toxicity in children. Using this to argue for the class on CDK6's rank, while the approved drugs hit both, is an argument for exposure to CDK4 blockade that this analysis explicitly does not support.
- Is a CDK6-selective inhibitor available, and would it separate the benefit from the CDK4-associated toxicity? Every approved agent in the class hits both.
- Do BET, SWI-SNF and CREBBP rank differently in lines made resistant to a menin inhibitor? That is the screen this analysis could not do and the one that would make their case.
- Does JNK dependency appear only once JUND is liberated? A drug-naive screen cannot see a conditional dependency, and this is the way the JNK negative could be wrong.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
In the 17,728-gene KMT2A-rearranged selectivity scan, four of the six proposed countermeasure targets clear the top 1% (rank ≤ 177): CDK6 rank 13, MCL1 18, BCL2 36, KAT7 92. Four do not: BRD4 3,122, MAPK8 4,220, SMARCA2 11,581, CREBBP 16,950. CDK4 ranks 16,052 against CDK6's 13, and SMARCA4 (16,333), CREBBP and CDK4 all sit on the less-required side.
Genome-wide KMT2A-rearranged selectivity scan, 17,728 genes (DepMap Chronos via Open Targets) · Broad Institute DepMap / Open Targets Platform; analyses/depmap-kmt2a-selective-scan-full 2026What this evidence cannot show
Computed by the coordinator on 2026-09-19 from an already-persisted scan; no new measurement. Eleven KMT2A-rearranged lines against 43 other leukaemia lines, none infant-derived. GENE EFFECT IS NOT DRUG EFFECT: a CDK4/6 inhibitor engages both genes whatever their individual ranks, and CRISPR knockout is not pharmacological inhibition. A difference statistic is least informative for strongly pan-essential genes, where a floor effect can compress or invert the sign - BRD4 sits at -1.01 pan-essentiality, which is exactly that regime.
Evidence that argues against this (1)
- This screens the wrong population for half the table. DepMap profiles UNSELECTED KMT2A-rearranged lines, and BET, SWI-SNF, CREBBP and CDK4/6 are all proposed for the state that exists AFTER menin-inhibitor exposure - MEN1-wild-type resistant cells, or cells that have lost KMT2C/KMT2D. A dependency acquired with resistance cannot appear in a screen of drug-naive lines, so a low rank there is an artefact of the design rather than a finding about the drug. Read the four failures as one informative negative (JNK) and three untested. Broad Institute DepMap / Open Targets Platform; analyses/depmap-kmt2a-selective-scan-full 2026
Built on the same comparator; CDK6 shows the largest shift and its gap from CDK4 widens.
- related to →JUND liberation — the drug's mechanism of action and one of its mechanisms of resistance are the same event
Its proposed countermeasure, JNK inhibition, is the only failure here not explained by the resistant-state design limitation: MAPK8 ranks 4,220 of 17,728.
Both its countermeasures clear the top 1% - MCL1 rank 18, BCL2 rank 36 - which is the strongest single-gene support any countermeasure in the beat has.
Its CDK4/6 recommendation averages two genes at opposite ends: CDK6 rank 13, CDK4 rank 16,052. The rationale should read CDK6.
SMARCA4 16,333, CREBBP 16,950, BRD4 3,122 in drug-naive lines - but this entry's claim is about the resistant state, which the scan cannot model, so the low ranks do not bear on it.
- supports →KAT7 / HBO1 — an enzyme, a leukaemia-stem-cell dependency, and an inhibitor already in a crystal
KAT7 ranks 92 of 17,728 and KAT6A 106, both inside the top 1%.
- supports →Experiment: time-to-resistance, not synergy index, as the readout for menin-inhibitor combinations
Supplies a cheap pre-ranking for the experiment that would actually compare these six head to head.
- related to →The whole protein-coding genome: menin is still first, and a metabolic pathway appears (run)
Consumes that scan's persisted per-gene ranks; no new computation.
The other cheap target-triage axis: this asks whether a target is selectively required, that one whether anyone has ever made a molecule that binds it.
- ← related toSix of 85 proteins are both selectively required and drugged — and MEN1 is not close to the others
Same selectivity axis, applied to proposed combination partners - and carrying the same objection, that single-agent lethality is the wrong test for a combination.