Ranking every KMT2A::AFF1 partner on dependency versus essentiality
also called therapeutic index ranking · partner selectivity scan · dependency-to-essentiality ratio
We scored every protein the fusion touches on one question: is it needed much more by KMT2A-rearranged leukemia cells than by cells in general? Menin wins, LEDGF and a barely-studied zinc finger called ZFP64 come next, and several famous partners turn out to be needed less by these cells than by other leukemias.
The same thing, for a scientist
For 41 candidate KMT2A::AFF1 interactors we pulled per-cell-line DepMap CRISPR (Chronos) gene effect through the Open Targets Platform GraphQL API, computed means over all 1,258 lines, over all leukemia/MPN lines, over a curated set of 10 KMT2A-rearranged lines and over the 4 KMT2A::AFF1 lines, and defined a lineage-controlled selectivity score = mean(KMT2A-r) minus mean(leukemia). Paired with gnomAD loss-of-function constraint this gives an explicit therapeutic-index ranking. MEN1 (-0.631) and KMT2A itself (-0.456) lead; PSIP1/LEDGF (-0.346) and ZFP64 (-0.318) are the next best and are far more tolerable genome-wide than the common-essential partners; WDR5, CTR9, AFF4, EAF2, ASH2L and RBBP5 score positively, meaning KMT2A-rearranged lines need them less than other leukemia lines do.
Tell me more about this picture
x axis: mean DepMap Chronos gene effect across 1,258 cell lines. y axis: mean Chronos in 10 KMT2A-rearranged lines minus mean Chronos in all leukemia lines.
Tell me more about this picture
MEN1 sits lowest on y. PSIP1 and ZFP64 sit near x = 0 and low on y, which is the combination the project is looking for.
Tell me more about this picture
CDK9, PRMT1, WDR5 and BRD4 sit far left, meaning every cell needs them; their y values near or above zero say the requirement is not KMT2A-specific.
Tell me more about this picture
With only 10 lines the y axis carries roughly plus or minus 0.15 of noise, so only the extremes should be read.
Why a lineage-controlled score, and what it is1 deeper
The naive comparison (KMT2A-rearranged lines against all 1,258 DepMap lines) confounds two things: genes that leukemia needs, and genes that this leukemia needs. Against a background dominated by solid-tumour lines, almost every chromatin and elongation factor looks selective.
So the score reported here is
selectivity_vs_leukemia = mean Chronos(KMT2A-r lines) - mean Chronos(all leukemia lines)
where the denominator is every line whose DepMap diseaseFromSource names a leukemia, a myeloproliferative neoplasm or an acute leukemia of ambiguous lineage (n = 70-77 per gene). A more negative score means a sharper KMT2A-rearranged-specific requirement over and above being a leukemia gene. The un-controlled score is reported alongside, and the gap between the two is itself informative 1.
The KMT2A-rearranged set is curated, not string-matched: MOLM-13, MOLM-14, THP-1, NOMO-1 (KMT2A::MLLT3), MV4-11, SEM, SEMK2, RS4;11 (KMT2A::AFF1), HB11;19 (KMT2A::MLLT1) and SHI-1 (KMT2A::AFDN). An earlier automated pass also matched EOL-1 and P31/FUJ; EOL-1 carries FIP1L1::PDGFRA and no KMT2A rearrangement and was dropped, and P31/FUJ was dropped as unconfirmed. KOPN-8 and ML-2 are not in this DepMap release. n = 10 is the honest sample size and it is small.
The ranking
Chronos gene effect; more negative = more required. sel(leuk) is the lineage-controlled score, sel(all) the naive one, CE the DepMap common-essential flag as served by Open Targets 1.
| # | Gene | sel(leuk) | sel(all) | all lines | leukemia | KMT2A-r | KMT2A::AFF1 | CE | LOEUF |
|---|---|---|---|---|---|---|---|---|---|
| 1 | MEN1 | -0.631 | -0.842 | -0.363 | -0.574 | -1.205 | -1.072 | no | 0.204 |
| 2 | KMT2A | -0.456 | -0.604 | -0.206 | -0.354 | -0.810 | -0.722 | no | 0.070 |
| 3 | PSIP1 | -0.346 | -0.472 | +0.028 | -0.098 | -0.444 | -0.346 | no | 0.373 |
| 4 | ZFP64 | -0.318 | -0.370 | -0.005 | -0.057 | -0.375 | -0.270 | no | 0.424 |
| 5 | KAT7 | -0.242 | -0.329 | -0.300 | -0.387 | -0.629 | -0.543 | no | 0.207 |
| 6 | AFF1 | -0.207 | -0.328 | +0.069 | -0.052 | -0.259 | -0.337 | no | 0.502 |
| 7 | ELL | -0.177 | -0.163 | -1.302 | -1.288 | -1.465 | -1.537 | yes | 0.417 |
| 8 | MEAF6 | -0.168 | -0.270 | -0.356 | -0.458 | -0.626 | -0.519 | no | 0.799 |
| 9 | DOT1L | -0.166 | -0.417 | -0.148 | -0.399 | -0.565 | -0.546 | no | 0.288 |
| 10 | MLLT1 (ENL) | -0.132 | -0.288 | -0.006 | -0.162 | -0.294 | -0.277 | no | 0.330 |
| 11 | LEO1 | -0.122 | -0.187 | -0.213 | -0.278 | -0.400 | -0.511 | no | 0.563 |
| 12 | BRPF1 | -0.120 | -0.244 | -0.262 | -0.386 | -0.506 | -0.404 | no | 0.201 |
| 13 | MED26 | -0.113 | +0.031 | -0.944 | -0.800 | -0.913 | -0.718 | yes | 0.301 |
| 14 | SETD1A | -0.100 | -0.043 | -0.911 | -0.854 | -0.954 | -0.990 | yes | 0.264 |
| 15 | MLLT10 | -0.084 | -0.127 | +0.061 | +0.018 | -0.066 | +0.102 | no | 0.315 |
| 16 | CDK9 | -0.077 | -0.151 | -1.536 | -1.610 | -1.687 | -1.552 | yes | 1.162 |
| 17 | DPF2 | -0.076 | -0.194 | -0.405 | -0.523 | -0.599 | -0.559 | no | 0.285 |
| 18 | CCNT1 | -0.055 | -0.047 | -0.206 | -0.198 | -0.253 | -0.047 | no | 0.397 |
| 19 | PAF1 | -0.046 | +0.008 | -0.832 | -0.778 | -0.824 | -0.929 | yes | 0.611 |
| 20 | BRD4 | -0.027 | -0.094 | -1.015 | -1.082 | -1.109 | -1.082 | yes | 0.171 |
| 21 | KDM2B | -0.020 | -0.087 | -0.040 | -0.107 | -0.127 | -0.155 | no | 0.240 |
| 22 | CDC73 | -0.014 | -0.075 | -1.122 | -1.183 | -1.197 | -1.189 | yes | 0.361 |
| 23 | ELL2 | -0.010 | +0.021 | -0.064 | -0.033 | -0.043 | +0.003 | no | 0.416 |
| 24 | ARID1A | -0.004 | -0.161 | -0.234 | -0.391 | -0.395 | -0.406 | no | 0.137 |
| 25 | PRMT1 | -0.003 | -0.180 | -1.399 | -1.576 | -1.579 | -1.583 | yes | 0.222 |
| 26 | SGF29 | 0.000 | -0.188 | -0.567 | -0.755 | -0.755 | -0.713 | no | 0.606 |
| 27 | JADE3 | +0.002 | -0.015 | -0.072 | -0.089 | -0.087 | -0.109 | no | 0.126 |
| 28 | SENP6 | +0.008 | +0.096 | -1.188 | -1.100 | -1.092 | -1.199 | yes | 0.396 |
| 29 | DPY30 | +0.010 | -0.050 | -0.390 | -0.450 | -0.440 | -0.439 | no | 0.731 |
| 30 | SMARCA4 | +0.019 | -0.123 | -0.364 | -0.506 | -0.487 | -0.485 | no | 0.216 |
| 31 | SIRT1 | +0.039 | +0.081 | +0.035 | +0.077 | +0.116 | +0.097 | no | 0.497 |
| 32 | CBX8 | +0.041 | -0.005 | +0.101 | +0.055 | +0.096 | +0.109 | no | 0.363 |
| 33 | MLLT3 (AF9) | +0.059 | +0.061 | -0.002 | 0.000 | +0.059 | +0.058 | no | 0.172 |
| 34 | EAF1 | +0.061 | +0.290 | -0.409 | -0.180 | -0.119 | -0.040 | no | 1.047 |
| 35 | RBBP5 | +0.063 | -0.036 | -1.105 | -1.204 | -1.141 | -1.226 | yes | 0.146 |
| 36 | BCOR | +0.063 | -0.022 | +0.025 | -0.060 | +0.003 | +0.048 | no | 0.196 |
| 37 | ASH2L | +0.093 | +0.159 | -0.333 | -0.267 | -0.174 | -0.177 | no | 0.502 |
| 38 | AFF4 | +0.111 | +0.082 | -0.056 | -0.085 | +0.026 | +0.024 | no | 0.166 |
| 39 | CTR9 | +0.112 | +0.022 | -0.873 | -0.963 | -0.851 | -0.908 | yes | 0.350 |
| 40 | EAF2 | +0.114 | +0.138 | -0.064 | -0.040 | +0.074 | +0.108 | no | 1.345 |
| 41 | WDR5 | +0.191 | +0.108 | -1.159 | -1.242 | -1.051 | -0.868 | yes | n/a |
What this says, including the parts that are inconvenient1 deeper
1. Menin's lead is not marginal. The gap between MEN1 (-0.631) and the next non-KMT2A gene (-0.346) is larger than the gap between third and twentieth place. On this metric the lab's central hypothesis survives its own test, and it survives against a lineage-matched control rather than against solid tumours.
2. LEDGF/PSIP1 is the best-positioned second interface in the table. Its mean Chronos across 1,258 lines is positive (+0.028) — knocking it out is on average neutral — and it still falls to -0.444 in KMT2A-rearranged lines. No other gene combines that degree of general dispensability with that size of selective requirement. If the project wants a second shot on goal with a better therapeutic index than menin, this is where the numbers point 1.
3. Four famous partners fail the test. WDR5 (+0.191), CTR9 (+0.112), AFF4 (+0.111) and RBBP5 (+0.063) are common-essential and required less by KMT2A-rearranged lines than by other leukemia lines. For WDR5 there is a structural reason rather than a data artefact: the WDR5-interaction (WIN) motif sits at KMT2A residues about 3762-3773, in the C-terminal half that KMT2A fusions delete, so WRAD is a wild-type-KMT2A module rather than a fusion partner 2. AFF4's failure is the sharper surprise, because AFF4 is usually drawn as the SEC scaffold; in KMT2A::AFF1 the fusion supplies its own AFF-family scaffold and the paralogue is not the dependency.
4. The common-essential band is a therapeutic-index wall, not a chemistry problem. ELL, CDK9, MED26, SETD1A, PAF1, BRD4, CDC73, PRMT1, RBBP5, CTR9 and SENP6 are all flagged common-essential with mean Chronos from -0.83 to -1.54. Several show the deepest absolute scores in KMT2A-rearranged lines — CDK9 at -1.687, PRMT1 at -1.579 — which is exactly how a target with no therapeutic window looks in a single-arm cell-line screen. Absolute depth of dependency is the wrong axis to rank on.
5. Two genes in the top eight are effectively undrugged and under-studied: ZFP64 (rank 4) and MEAF6 (rank 8). Each has its own hypothesis node.
Limits a referee would attack first
- n = 10 KMT2A-rearranged lines, 4 of them KMT2A::AFF1. At that sample size a selectivity score of -0.1 is noise. Only roughly the top five entries are likely to survive dropping any single line, and we have not bootstrapped this.
- The means were computed by us from per-line values served by the Open Targets API, not taken from a DepMap release summary table; the release version is not stated in the API response.
- Cell lines are neither patients nor normal tissue. A gene can be dispensable across 1,258 immortalised lines and still be required by a human haematopoietic stem cell under stress, which is exactly what happened with menin (see
gene-men1). - CRISPR knockout is not pharmacological inhibition of one interface. PSIP1's score reflects loss of the whole protein, removing the PWWP domain, the IBD and everything else at once. A drug blocking only the IBD-menin contact would do less.
- gnomAD LOEUF and DepMap disagree constructively and must not be averaged. ZFP64 has pLI ~1.0 and LOEUF 0.424 yet a mean Chronos of -0.005; constraint measures multi-decade selection on heterozygotes, not cell viability.
- No expression filter was applied. A gene not expressed in a line cannot score as a dependency there, which flattens scores for tissue-restricted partners.
How to re-run this
Three API calls per gene, no local data:
- Resolve the symbol to an Ensembl id with the Open Targets
searchquery. target(ensemblId:){ isEssential depMapEssentiality { tissueName screens { cellLineName diseaseFromSource geneEffect } } }, then aggregate over the curated line sets, de-duplicating oncellLineName|tissueNamebecause some lines appear under more than one tissue.- gnomAD GraphQL
gene(gene_symbol:, reference_genome:GRCh38){ gnomad_constraint{ pLI oe_lof_upper obs_lof exp_lof mis_z } }.
Two practical traps worth recording: the AlphaFold and gnomAD APIs reject Node's default user agent with HTTP 403 and need a browser-like User-Agent; and the DepMap portal is a client-side application that cannot be scraped, so the Open Targets API is the way in.
| script | analyses/partner-dependency-vs-essentiality/run.mjs |
|---|---|
| status | run |
| method | selectivity_vs_leukemia = mean Chronos(curated KMT2A-rearranged lines) - mean Chronos(all leukemia/MPN lines); per-line values de-duplicated on cellLineName|tissueName |
| n genes | 41 |
| n depmap lines total | 1258 |
| n kmt2a rearranged lines | 10 |
| n kmt2a aff1 lines | 4 |
| result summary | MEN1 has the best dependency-to-essentiality profile of any KMT2A::AFF1 partner (selectivity -0.631, not common-essential). PSIP1/LEDGF is the best non-menin candidate: mean Chronos +0.028 across 1,258 lines yet -0.444 in KMT2A-rearranged lines. ZFP64 ranks fourth and is essentially undrugged. WDR5, CTR9, AFF4, EAF2, ASH2L and RBBP5 are anti-selective, i.e. KMT2A-rearranged lines need them less than other leukemia lines do. |
inputs
[ "Open Targets Platform GraphQL API: target.depMapEssentiality, target.isEssential", "gnomAD GraphQL API: gene.gnomad_constraint", "Human Protein Atlas per-gene JSON", "RCSB PDB search and data GraphQL APIs" ]
kmt2a rearranged lines
[
{
"line": "MOLM-13",
"fusion": "KMT2A::MLLT3"
},
{
"line": "MOLM-14",
"fusion": "KMT2A::MLLT3"
},
{
"line": "THP-1",
"fusion": "KMT2A::MLLT3"
},
{
"line": "NOMO-1",
"fusion": "KMT2A::MLLT3"
},
{
"line": "MV4-11",
"fusion": "KMT2A::AFF1"
},
{
"line": "SEM",
"fusion": "KMT2A::AFF1"
},
{
"line": "SEMK2",
"fusion": "KMT2A::AFF1"
},
{
"line": "RS4;11",
"fusion": "KMT2A::AFF1"
},
{
"line": "HB11;19",
"fusion": "KMT2A::MLLT1"
},
{
"line": "SHI-1",
"fusion": "KMT2A::AFDN"
}
]lines excluded
[
{
"line": "EOL-1",
"reason": "carries FIP1L1::PDGFRA, not a KMT2A rearrangement"
},
{
"line": "P31/FUJ",
"reason": "KMT2A rearrangement status not confirmed to our satisfaction"
},
{
"line": "KOPN-8",
"reason": "not present in this DepMap release"
},
{
"line": "ML-2",
"reason": "not present in this DepMap release"
}
]ranking columns
[ "rank", "gene", "selectivity_vs_leukemia", "selectivity_vs_all_lines", "chronos_mean_all", "chronos_mean_leukemia", "chronos_mean_kmt2ar", "chronos_mean_kmt2a_aff1", "depmap_common_essential", "gnomad_pLI", "gnomad_loeuf" ]
ranking
[
[
1,
"MEN1",
-0.631,
-0.842,
-0.363,
-0.574,
-1.205,
-1.072,
false,
1,
0.204
],
[
2,
"KMT2A",
-0.456,
-0.604,
-0.206,
-0.354,
-0.81,
-0.722,
false,
1,
0.07
],
[
3,
"PSIP1",
-0.346,
-0.472,
0.028,
-0.098,
-0.444,
-0.346,
false,
1,
0.373
],
[
4,
"ZFP64",
-0.318,
-0.37,
-0.005,
-0.057,
-0.375,
-0.27,
false,
0.9996,
0.424
],
[
5,
"KAT7",
-0.242,
-0.329,
-0.3,
-0.387,
-0.629,
-0.543,
false,
1,
0.207
],
[
6,
"AFF1",
-0.207,
-0.328,
0.069,
-0.052,
-0.259,
-0.337,
false,
0.7645,
0.502
],
[
7,
"ELL",
-0.177,
-0.163,
-1.302,
-1.288,
-1.465,
-1.537,
true,
0.9999,
0.417
],
[
8,
"MEAF6",
-0.168,
-0.27,
-0.356,
-0.458,
-0.626,
-0.519,
false,
0.0095,
0.799
],
[
9,
"DOT1L",
-0.166,
-0.417,
-0.148,
-0.399,
-0.565,
-0.546,
false,
1,
0.288
],
[
10,
"MLLT1",
-0.132,
-0.288,
-0.006,
-0.162,
-0.294,
-0.277,
false,
1,
0.33
],
[
11,
"LEO1",
-0.122,
-0.187,
-0.213,
-0.278,
-0.4,
-0.511,
false,
0.0966,
0.563
],
[
12,
"BRPF1",
-0.12,
-0.244,
-0.262,
-0.386,
-0.506,
-0.404,
false,
1,
0.201
],
[
13,
"MED26",
-0.113,
0.031,
-0.944,
-0.8,
-0.913,
-0.718,
true,
1,
0.301
],
[
14,
"SETD1A",
-0.1,
-0.043,
-0.911,
-0.854,
-0.954,
-0.99,
true,
1,
0.264
],
[
15,
"MLLT10",
-0.084,
-0.127,
0.061,
0.018,
-0.066,
0.102,
false,
1,
0.315
],
[
16,
"CDK9",
-0.077,
-0.151,
-1.536,
-1.61,
-1.687,
-1.552,
true,
0,
1.162
],
[
17,
"DPF2",
-0.076,
-0.194,
-0.405,
-0.523,
-0.599,
-0.559,
false,
1,
0.285
],
[
18,
"CCNT1",
-0.055,
-0.047,
-0.206,
-0.198,
-0.253,
-0.047,
false,
1,
0.397
],
[
19,
"PAF1",
-0.046,
0.008,
-0.832,
-0.778,
-0.824,
-0.929,
true,
0.0106,
0.611
],
[
20,
"BRD4",
-0.027,
-0.094,
-1.015,
-1.082,
-1.109,
-1.082,
true,
1,
0.171
],
[
21,
"KDM2B",
-0.02,
-0.087,
-0.04,
-0.107,
-0.127,
-0.155,
false,
1,
0.24
],
[
22,
"CDC73",
-0.014,
-0.075,
-1.122,
-1.183,
-1.197,
-1.189,
true,
1,
0.361
],
[
23,
"ELL2",
-0.01,
0.021,
-0.064,
-0.033,
-0.043,
0.003,
false,
0.9999,
0.416
],
[
24,
"ARID1A",
-0.004,
-0.161,
-0.234,
-0.391,
-0.395,
-0.406,
false,
1,
0.137
],
[
25,
"PRMT1",
-0.003,
-0.18,
-1.399,
-1.576,
-1.579,
-1.583,
true,
1,
0.222
],
[
26,
"SGF29",
0,
-0.188,
-0.567,
-0.755,
-0.755,
-0.713,
false,
0.4902,
0.606
],
[
27,
"JADE3",
0.002,
-0.015,
-0.072,
-0.089,
-0.087,
-0.109,
false,
1,
0.126
],
[
28,
"SENP6",
0.008,
0.096,
-1.188,
-1.1,
-1.092,
-1.199,
true,
1,
0.396
],
[
29,
"DPY30",
0.01,
-0.05,
-0.39,
-0.45,
-0.44,
-0.439,
false,
0.5125,
0.731
],
[
30,
"SMARCA4",
0.019,
-0.123,
-0.364,
-0.506,
-0.487,
-0.485,
false,
1,
0.216
],
[
31,
"SIRT1",
0.039,
0.081,
0.035,
0.077,
0.116,
0.097,
false,
0.9779,
0.497
],
[
32,
"CBX8",
0.041,
-0.005,
0.101,
0.055,
0.096,
0.109,
false,
0.9999,
0.363
],
[
33,
"MLLT3",
0.059,
0.061,
-0.002,
0,
0.059,
0.058,
false,
1,
0.172
],
[
34,
"EAF1",
0.061,
0.29,
-0.409,
-0.18,
-0.119,
-0.04,
false,
0,
1.047
],
[
35,
"RBBP5",
0.063,
-0.036,
-1.105,
-1.204,
-1.141,
-1.226,
true,
1,
0.146
],
[
36,
"BCOR",
0.063,
-0.022,
0.025,
-0.06,
0.003,
0.048,
false,
1,
0.196
],
[
37,
"ASH2L",
0.093,
0.159,
-0.333,
-0.267,
-0.174,
-0.177,
false,
0.9535,
0.502
],
[
38,
"AFF4",
0.111,
0.082,
-0.056,
-0.085,
0.026,
0.024,
false,
1,
0.166
],
[
39,
"CTR9",
0.112,
0.022,
-0.873,
-0.963,
-0.851,
-0.908,
true,
1,
0.35
],
[
40,
"EAF2",
0.114,
0.138,
-0.064,
-0.04,
0.074,
0.108,
false,
0,
1.345
],
[
41,
"WDR5",
0.191,
0.108,
-1.159,
-1.242,
-1.051,
-0.868,
true,
null,
null
]
]top candidates for phase2
[ "gene-psip1", "gene-zfp64", "gene-kat7", "gene-dot1l", "gene-mllt1" ]
anti selective partners
[ "WDR5", "CTR9", "AFF4", "EAF2", "ASH2L", "RBBP5", "MLLT3", "CBX8", "BCOR", "SIRT1" ]
- Ranking on cell-line data systematically under-weights toxicity in non-dividing and stem-cell compartments; menin's own long-term HSC defect is invisible to this metric.
- A high selectivity score for a multi-domain protein such as PSIP1 or KAT7 does not transfer to a drug that blocks only one interface; the score is an upper bound on what chemistry can achieve.
- The anti-selective genes are still lethal when knocked out in KMT2A-rearranged cells. A programme that read 'not selective' as 'will not work' would discard active agents; the claim is about therapeutic index, not activity.
- If the fusion can switch between redundant partners (AFF1 and AFF4, ELL and ELL2, ENL and AF9), single-gene CRISPR scores will systematically understate every member of a redundant pair, and the whole ranking is biased against exactly the nodes worth drugging in combination.
- With only 10 KMT2A-rearranged lines, is the PSIP1 and ZFP64 signal stable under leave-one-out and under a matched permutation null? This is a one-afternoon analysis and it decides whether ranks 3 and 4 are real.
- Does the ranking change if the denominator is restricted to KMT2A-wild-type B-ALL and AML lines rather than all leukemia, and does KMT2A::AFF1 differ from KMT2A::MLLT3 in which partners it needs?
- Which of these selectivity scores survive when the perturbation is interface-specific, for example a LEDGF IBD point mutant rather than a PSIP1 knockout?
- None of these lines is an infant KMT2A::AFF1 primary sample. Does the ranking reproduce in primary infant ALL blasts or patient-derived xenografts, where the cell of origin is fetal rather than immortalised?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
Per-cell-line DepMap CRISPR (Chronos) gene effect, the common-essential flag, gnomAD loss-of-function constraint and Human Protein Atlas expression for 41 candidate KMT2A::AFF1 interactors, from which the lineage-controlled selectivity scores in this node were computed.
DepMap CRISPR (Chronos) gene effect and essentiality flags retrieved through the Open Targets Platform GraphQL API; loss-of-function constraint from the gnomAD GraphQL API; normal-tissue expression from the Human Protein Atlas API · Broad Institute DepMap; gnomAD; Human Protein Atlas; Open Targets Platform · Databases 2026What this evidence cannot show
Means and selectivity scores were computed by us from per-line API values rather than read from a DepMap release table, and the release version is not reported by the API. The KMT2A-rearranged set is only 10 lines (4 KMT2A::AFF1), so scores under about 0.15 in absolute value should be treated as noise. Cell-line essentiality is a poor model of normal-tissue tolerance, and whole-gene CRISPR knockout is not equivalent to blocking one protein-protein interface. WDR5 returned no gnomAD constraint record.
Evidence that argues against this (1)
- The strongest argument against ranking on cell-line data is menin itself: MEN1 has a modest mean Chronos yet conditional deletion causes a marked long-term haematopoietic stem cell defect under competitive transplantation, a toxicity no cell-line screen can see. Maillard I et al. 2009
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
The KMT2A WDR5-interaction (WIN) motif lies at approximately KMT2A residues 3762-3773, inside the C-terminal region that KMT2A fusions delete, which is the structural reason WRAD-module genes do not score as KMT2A-rearrangement-selective dependencies.
WDR5 bound to MLL1 peptide (PDB 3EG6, KMT2A 3762-3773) and structural basis of WDR5-MLL interaction (PDB 3EMH, KMT2A 3764-3776); MLL1 core complex on the nucleosome (PDB 6KIU and 6PWV, KMT2A 3754-3969 and 3762-3969) - per-entity UniProt residue ranges resolved against the RCSB PDB data API · RCSB Protein Data Bank · RCSB PDB 2026What this evidence cannot show
The residue ranges quoted are the crystallographically modelled KMT2A segments as annotated by RCSB against UniProt Q03164, not an independent mapping of the WIN motif; reported motif boundaries differ by a residue or two between papers. Which KMT2A residues a patient's fusion retains depends on the breakpoint, so this is a statement about the common intron 9-11 breakpoints, not a universal one.
Evidence that argues against this (1)
- WDR5 remains a deep dependency in absolute terms in KMT2A-rearranged lines (mean Chronos -1.051), so a WDR5 inhibitor would kill these cells. The claim here is only that it would not be selective for them, and the common-essential flag predicts on-target toxicity elsewhere. Broad Institute DepMap; Open Targets Platform 2026
- ← depends onExperiment: combinatorial dose-titration of the HOXA9/MEIS1 program to find its minimal cut set
node selection should follow the existing dependency-versus-essentiality ranking
- ← related toRUNX1::RUNX1T1 (AML1-ETO), t(8;21)(q22;q22)
Essentiality confound (Goyama 2013, PMID 23979164): RUNX1 is also required for MLL-AF9 cell growth and for MLL-AF9 leukaemia development in mice, so native-RUNX1 dependence is not a t(8;21)-selective vulnerability.
ranks every partner on required-by-the-cancer against needed-by-normal-tissue
reaches the same conclusion from correlation that the other analysis reaches from essentiality