ZFP64 — rank 4 on selectivity, and there is not even a ChEMBL entry for it
also called ZFP64 · zinc finger protein 64 · ZNF338
ZFP64 is a zinc-finger protein that sits on the KMT2A gene's own promoter and keeps KMT2A switched on. Cells in general do not need it, KMT2A-rearranged leukemia cells do, and no drug programme anywhere is working on it.
The same thing, for a scientist
ZFP64 (Q9NTW7, 645 aa) is a C2H2 zinc-finger transcription factor identified by a domain-focused CRISPR screen as essential in KMT2A-rearranged leukemia. Its critical function is sustaining KMT2A gene expression by occupying the KMT2A promoter, which carries an exceptional density of ZFP64 motifs — a transcription-factor addiction imposed by promoter sequence rather than a direct protein-protein contact with the fusion. Its DepMap profile is one of the best in the interactome: mean Chronos -0.005 across 1,258 lines (neutral), -0.375 across 10 KMT2A-rearranged lines, lineage-controlled selectivity -0.318, rank 4 behind only MEN1, KMT2A and PSIP1. It has no ChEMBL target entity at all and only two isolated NMR zinc-finger structures.
Tell me more about this picture
DepMap Chronos gene effect; 0 is no effect and more negative means the cells need the gene more.
Tell me more about this picture
The step from -0.057 in leukemia generally to -0.375 in KMT2A-rearranged lines is the lineage-controlled selectivity of -0.318 that puts ZFP64 fourth of 41 partners.
Tell me more about this picture
None of the context means crosses the conventional -0.5 depletion threshold, so this is a graded requirement rather than an absolute one - which is consistent with a mechanism that lowers fusion dose rather than abolishing fusion function.
Tell me more about this picture
With only 10 KMT2A-rearranged lines this comparison has not been leave-one-out tested; that test is the first thing to run.
This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.
I have never read one of these pictures before
A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.
Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.
Anything in yellow sticks is a bound small molecule, which is usually the drug.
Why this node exists1 deeper
ZFP64 was not on the original list of things to map. It came out of the dependency-versus-essentiality scan (ana-partner-dependency-vs-essentiality) and then turned out to have a published mechanism that nobody in the menin-inhibitor conversation talks about.
The numbers 1:
| value | |
|---|---|
| mean Chronos, 1,258 lines | -0.005 (neutral) |
| fraction of lines depleted | 16.5% |
| mean Chronos, all leukemia lines | -0.057 |
| mean Chronos, 10 KMT2A-rearranged lines | -0.375 |
| mean Chronos, 4 KMT2A::AFF1 lines | -0.270 |
| selectivity vs leukemia | -0.318 (rank 4 of 41) |
| DepMap common-essential | no |
| gnomAD pLI / LOEUF | 0.9996 / 0.424 |
| HPA | low tissue specificity, detected in all tissues |
That profile is what the project says it is looking for: near-total dispensability in the average cell, a real requirement in the right cell, and no common-essential flag. Only menin, KMT2A itself and LEDGF score better.
The mechanism is upstream, not a protein-protein contact — and that changes the target logic
The published mechanism is specific and it is not what the framing of this beat assumes. A domain-focused CRISPR screen identified ZFP64 as essential in KMT2A-rearranged leukemia, and the critical function is sustaining expression of the KMT2A gene by binding the KMT2A promoter, which is its single most enriched genomic occupancy and carries an exceptional density of ZFP64 motifs. The authors frame it as a transcription-factor addiction imposed by promoter sequence 2.
So ZFP64 is not an interactome member in the sense that menin or LEDGF is. It does not grip the fusion protein. It keeps the fusion protein's gene transcribed. We could find no verified evidence that ZFP64 physically contacts KMT2A protein or recruits it to chromatin, and this node makes no such claim.
That distinction has a real therapeutic consequence, and it cuts in ZFP64's favour. Every interface target in this dossier — menin, LEDGF, the AF9 AHD — attacks the fusion's function, which selects for resistance mutations that restore function while excluding the drug (exactly what MEN1 interface mutations do 3). Attacking the fusion's expression is a different axis. A cell cannot mutate its way out of needing the fusion transcript, only out of needing ZFP64, and it would have to do that without losing fusion expression.
Why nobody is drugging it, and whether that is a good reason
There is no ZFP64 target entity in ChEMBL v34 at all. Searching by gene symbol with and without organism, and by the target name 'Zinc finger protein 64', returns zero records 4. Not an empty target — no target.
Structurally it is almost as bare. Two PDB entries exist, both isolated NMR zinc-finger domains from structural-genomics deposition: 1X5W (ZFP64 residues 493-548) and 2DMD (ZFP64 residues 174-254). There is no ZFP64-DNA structure and no ZFP64-partner structure 5. An AlphaFold model exists (AF-Q9NTW7-F1) and resolves, which is more than can be said for full-length KMT2A.
The honest reading is that C2H2 zinc-finger transcription factors are a class the small-molecule field has largely written off: the DNA-binding surface is extended and basic, there is no enzymatic activity, and selectivity across a family of hundreds of homologous fingers is a serious problem. That is a defensible reason for the gap, not an oversight.
But the modality landscape has changed since that judgement was formed. Three routes are now realistic and none requires a classical pocket:
- Targeted degradation. ZFP64 is a nuclear protein of modest size with no essential enzymatic function to preserve; a molecular-glue or PROTAC approach needs only a bindable surface, not a functional pocket. The precedent that this is feasible against a chromatin reader in this exact disease is
9DUR, a CRBN-degrader-ENL YEATS ternary complex solved at 2.90 Å. - Blocking the promoter, not the protein. The mechanism is motif-density-driven occupancy of one promoter. That is addressable by oligonucleotide or CRISPRi approaches in a way a diffuse protein-protein interface is not.
- Synthetic lethality. If ZFP64 loss lowers fusion dose rather than abolishing it, then ZFP64 partial inhibition plus a menin inhibitor should be strongly synergistic — lower fusion dose means less clamp to displace.
What would make me drop this candidate
Three results would kill it, and all three are cheap.
First, leave-one-out instability. The selectivity score rests on 10 cell lines. If dropping MOLM-13 or SEMK2 moves -0.318 to -0.15, the ranking position is an artefact. This has not been tested and it is an afternoon's work.
Second, if the dependency is not KMT2A-specific at all. The mechanism as published is maintenance of KMT2A expression. Wild-type KMT2A is also expressed from the same promoter in every cell, and the DepMap score for KMT2A itself in non-KMT2A-rearranged leukemia lines is -0.354, i.e. non-trivial. If ZFP64's apparent selectivity simply tracks how much any given line depends on KMT2A dosage, then ZFP64 is a KMT2A-dosage target rather than a fusion target, and its therapeutic index collapses toward KMT2A's own.
Third, gnomAD. pLI 0.9996 with LOEUF 0.424 and 14 observed loss-of-function variants against 51.6 expected means human heterozygous loss is selected against 1. As with menin and LEDGF this is a constraint signature rather than a viability signature, but the pattern is now consistent across the three best-scoring candidates in this dossier, and treating it as uninformative three times in a row would be motivated reasoning.
| uniprot | Q9NTW7 |
|---|---|
| gene symbol | ZFP64 |
| length AA | 645 |
| protein name | Zinc finger protein 64 |
| alphafold ID | AF-Q9NTW7-F1 |
| alphafold verified | true |
| PDB verified against rcsb | true |
| structural gap | no ZFP64-DNA structure, no ZFP64-partner structure, no structure of more than one zinc finger at a time |
| mechanism class | transcription-factor addiction: maintains KMT2A gene expression by occupying the motif-dense KMT2A promoter; NOT a demonstrated physical interactor of KMT2A protein |
| planned hypotheses note | Moved out of a reference field on 2026-09-19. It named a node that does not exist, which made the pointer read as a real cross-reference and, for hypotheses, silently satisfied the 'needs at least one experiment or analysis' rule. The intent is preserved here; the node still needs writing. |
PDB IDS
[ "1X5W", "2DMD" ]
key structures
[
{
"pdb": "1X5W",
"what": "isolated C2H2 zinc-binding domain, ZFP64 493-548",
"method": "NMR",
"resolution_A": null
},
{
"pdb": "2DMD",
"what": "isolated N-terminal C2H2 zinc-binding domain, ZFP64 174-254",
"method": "NMR",
"resolution_A": null
}
]essentiality
{
"depmap_common_essential": false,
"chronos_mean": -0.005,
"n_lines": 1258,
"fraction_lines_depleted": 0.165,
"chronos_mean_leukemia": -0.057,
"chronos_mean_kmt2ar": -0.375,
"chronos_mean_kmt2a_aff1": -0.27,
"selectivity_vs_leukemia": -0.318,
"selectivity_vs_all_lines": -0.37,
"rank_in_interactome": 4,
"gnomad_pLI": 0.9996,
"gnomad_loeuf": 0.424,
"gnomad_obs_lof": 14,
"gnomad_exp_lof": 51.6,
"gnomad_mis_z": 2.16,
"mouse_ko": "not established in this node; no verified citation obtained this cycle",
"hpa_rna_tissue_specificity": "Low tissue specificity",
"hpa_rna_tissue_distribution": "Detected in all",
"hpa_rna_blood_cell_specificity": "Low immune cell specificity"
}chemical matter
{
"chembl_target": null,
"chembl_status": "no ChEMBL target entity exists for ZFP64 in v34",
"n_activity_rows_pchembl_ge_6": 0,
"curated_mechanism_drugs": [],
"note": "not an empty target record - no target record at all. C2H2 zinc-finger transcription factors are a class classical small-molecule discovery has largely avoided."
}proposed modalities
[ "targeted degradation (molecular glue or PROTAC), since no enzymatic function must be preserved; feasibility precedent in this disease is the CRBN-degrader-ENL YEATS ternary complex PDB 9DUR at 2.90 A", "oligonucleotide or CRISPRi interference at the motif-dense KMT2A promoter rather than at the protein", "combination with a menin inhibitor on the hypothesis that lowering fusion dose reduces the amount of clamp the inhibitor must displace" ]
planned hypotheses
[ "hyp-zfp64-dose-lowering-combination" ]
depmap kmt2a selectivity
{
"chronos_mean_kmt2a_rearranged": -0.3919,
"chronos_mean_other_leukaemia": 0.0179,
"difference": -0.4098,
"pan_essentiality_all_lines": -0.0051,
"selectivity_rank": 32,
"rank_denominator": 17728,
"provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
"how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}- Essential-function risk hidden by cell lines: gnomAD pLI 0.9996 and LOEUF 0.424 say human heterozygous loss is selected against. The reassuring DepMap number comes from immortalised cancer lines and cannot see stem-cell or developmental requirements.
- Compensation by paralogous zinc fingers: if another C2H2 factor can occupy the motif-dense KMT2A promoter, a ZFP64 inhibitor would be buffered in a way a single-gene CRISPR knockout does not reveal.
- Resistance by promoter mutation: an addiction imposed by promoter sequence can in principle be escaped by altering that sequence or by enhancer hijacking that supplies KMT2A expression from elsewhere.
- On-target toxicity is completely uncharacterised. There is no mouse knockout phenotype in this dossier, no chemical probe, and no tissue in which ZFP64 is absent - so the entire toxicity argument currently rests on a DepMap mean of -0.005.
- Modality risk: all three routes proposed here (degrader, oligonucleotide, CRISPRi) carry delivery problems in infants that are harder than the target biology, and none has a starting molecule.
- Is the selectivity score stable? With 10 KMT2A-rearranged lines, leave-one-out and a permutation null are the first tests and they have not been run. If -0.318 collapses when MOLM-13 or SEMK2 is dropped, the rank-4 position is an artefact.
- Is ZFP64 a fusion target or a KMT2A-dosage target? Its published mechanism is maintenance of expression from a promoter shared with wild-type KMT2A. Comparing ZFP64 dependency against KMT2A dependency line by line would separate the two, and if they track together the therapeutic index collapses toward KMT2A's own.
- Does ZFP64 knockdown lower fusion transcript dose, and is that synergistic with a menin inhibitor? A simple two-drug matrix in SEM and MV4-11 with fusion transcript quantification would answer it.
- Does ZFP64 physically contact the KMT2A fusion protein at all? No verified evidence exists either way, and an immunoprecipitation in a KMT2A::AFF1 line would settle whether this node belongs in the interactome at all or only upstream of it.
- Is ZFP64 degradable? A ligandability assessment (thermal shift or a covalent fragment screen against recombinant ZFP64) is the gating experiment for every modality proposed here, and no one has published one.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
ZFP64 is not DepMap common-essential and has a mean Chronos gene effect of -0.005 across 1,258 cell lines with 16.5% of lines depleted, yet falls to -0.375 across 10 KMT2A-rearranged lines and -0.270 across 4 KMT2A::AFF1 lines, giving a lineage-controlled selectivity of -0.318 and rank 4 of 41 candidate partners; gnomAD gives pLI 0.9996, LOEUF 0.424, 14 observed versus 51.6 expected loss-of-function variants and mis_z 2.16; HPA classifies it as low tissue specificity, detected in all tissues, low immune cell specificity.
DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL API; loss-of-function constraint via the gnomAD GraphQL API; normal-tissue expression via the Human Protein Atlas API · Broad Institute DepMap; gnomAD; Human Protein Atlas; Open Targets Platform · Databases 2026What this evidence cannot show
Means were computed by us from per-line API values rather than read from a DepMap release table, and the release version is not stated by the API. Only 10 KMT2A-rearranged lines (4 KMT2A::AFF1) contribute, so the selective score has not been bootstrapped or leave-one-out tested and a value of -0.318 could move substantially. No expression filter was applied. Cell-line essentiality says nothing about normal haematopoietic stem cells.
Evidence that argues against this (1)
- gnomAD constraint points the other way: pLI 0.9996 and LOEUF 0.424 mean human heterozygous loss of ZFP64 is selected against, so tissue-wide inhibition is not obviously benign. The same tension recurs for MEN1 and PSIP1. gnomAD Consortium 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
A domain-focused CRISPR screen identifies ZFP64 as essential in KMT2A-rearranged leukemia, and its critical function is sustaining KMT2A gene expression by binding the KMT2A promoter, which is its most enriched genomic occupancy and carries an exceptional density of ZFP64 motifs - a transcription-factor addiction imposed by promoter sequence.
A transcription factor addiction in leukemia imposed by the MLL promoter sequence · Lu B, Klingbeil O, Tarumoto Y, et al. · Cancer Cell 2018What this evidence cannot show
The mechanism is transcriptional maintenance of the KMT2A gene, NOT recruitment of KMT2A protein to chromatin and NOT a physical ZFP64-fusion interaction; this node makes no such claim and no verified evidence for one was found. The work is a cell-line CRISPR screen with follow-up, not an in vivo therapeutic study, and no ZFP64 chemical probe was used. Because the promoter is shared with wild-type KMT2A, the result does not by itself establish that the dependency is specific to the fusion rather than to KMT2A dosage.
Evidence that argues against this (1)
- Sustaining KMT2A gene EXPRESSION by binding the KMT2A promoter is an indirect mechanism with an awkward consequence: the promoter is shared by the wild-type allele and the fusion allele, so ZFP64 loss lowers both, and the therapeutic reading depends on which matters more. A domain-focused CRISPR screen also finds what its library covers. Oncofusion knowledge graph (internal analysis) 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MEN1 interface mutations cause clinical resistance to menin inhibition by attenuating drug binding while preserving the menin-KMT2A interaction, illustrating that interface-directed agents select for function-restoring resistance - the failure mode an expression-directed target such as ZFP64 would not share.
MEN1 mutations mediate clinical resistance to menin inhibition · Perner F, Stein EM, Wenge DV, et al. · Nature 2023What this evidence cannot show
The comparison between interface-directed and expression-directed resistance is our inference, not a result in this paper. No ZFP64-directed agent exists, so the claim that it would resist differently is untested. Resistance to loss of a transcription factor can also arise, for example by promoter mutation or by compensating paralogues.
Evidence that argues against this (1)
- Interface mutations selecting for function-restoring change is a general lesson from MEN1 and it is cited here without a ZFP64 equivalent. No resistance mutation to any ZFP64-directed agent exists, because no such agent exists. Nature 2023
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
ChEMBL v34 contains no ZFP64 target entity whatsoever: searches by gene symbol with and without organism restriction, and by the target name 'Zinc finger protein 64', all return zero records.
ChEMBL database version 34: target_search queries for gene_symbol ZFP64 and target_name 'Zinc finger protein 64', with and without organism restriction, all returning zero results · EMBL-EBI ChEMBL · ChEMBL 2026What this evidence cannot show
A null result in ChEMBL is evidence of absent curated medicinal chemistry, not proof that no compound has ever been reported; patent-only or unpublished matter would not appear. ZFP64 shares the alias ZNF338, and while the searches performed should have caught it, an alias-driven miss cannot be excluded entirely.
Evidence that argues against this (1)
- A ChEMBL activity-row count measures how much medicinal chemistry has been DONE, not how tractable the target is, and the best reported affinity belongs to the most optimised compound in whichever series someone chose to pursue. Zero rows means nobody tried; thousands of rows means a target had a decade of industrial attention. Zero records across every query formulation is the cleanest possible statement that nobody has tried, and it says nothing about whether a zinc-finger transcription factor could be drugged. ChEMBL, EMBL-EBI 2026
- E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
ZFP64 has only two PDB entries, both isolated NMR zinc-finger domains (1X5W, residues 493-548; 2DMD, residues 174-254); there is no ZFP64-DNA structure and no ZFP64-partner structure. An AlphaFold model AF-Q9NTW7-F1 exists and resolves, unlike full-length KMT2A, for which AF-Q03164-F1 returns HTTP 404.
RCSB PDB entries 1X5W and 2DMD for UniProt Q9NTW7, and AlphaFold Protein Structure Database entry AF-Q9NTW7-F1, all resolved directly by API · RCSB Protein Data Bank; AlphaFold Protein Structure Database · Databases 2026What this evidence cannot show
The two NMR entries are structural-genomics depositions of single domains and carry no functional or ligand information. The AlphaFold model was confirmed only to exist and return HTTP 200; its per-residue confidence was not assessed, and for a protein dominated by tandem zinc fingers separated by linkers the inter-domain arrangement should be assumed unreliable.
Evidence that argues against this (1)
- A solution NMR structure of an isolated domain is the weakest structural evidence in this dossier class: the construct is chosen to be soluble and monomeric, the interaction partners are absent, and the model is an ensemble rather than a density map. Two isolated zinc fingers out of a protein with many, no DNA complex and no partner complex, is the structural profile of a protein that has been touched rather than studied. RCSB PDB 2026
- regulates →KMT2A / MLL1 — the gene that breaks
rediscovered at rank 12 with no prior; ZFP64 maintains MLL expression by binding the MLL promoter