SGK1 / SGK (O00141) - a MENIN-SELECTIVE upregulated kinase, a druggable AGC family member, and the panel's worst platform disagreement
also called SGK1 · SGK · SGK1 kinase · Serine/threonine-protein kinase Sgk1 · SGK1_HUMAN
SGK1 is a stress-activated kinase closely related to AKT that controls ion channels, transporters and survival signalling. It is one of the very few genes in this signature that goes UP specifically under a menin-blocking drug and not under a chromatin-enzyme-blocking one, and it is the kind of enzyme drugs can be made against.
The same thing, for a scientist
SGK1 (O00141, 431 aa, protein-kinase domain 98-355 plus an AGC-kinase C-terminal region 356-431) is a stress-responsive AGC-family serine/threonine kinase regulating ion channels, membrane transporters, transcription factors, growth, survival, migration and apoptosis, with a subcellular distribution controlled by the cell cycle, by hormones and by environmental stress. It is one of the **eight concordantly UP genes** and one of only **eight MENIN-INHIBITOR-SELECTIVE genes** in this beat's 37-gene specificity table: VTP-50469 gives **+1.99** in RS4;11 at day 7 while EPZ5676 gives **+0.07** - a **1.92 log2 difference, the largest menin-versus-DOT1L gap of any UP gene in the panel** [E1]. **But it is also the panel's worst platform disagreement, and that has to come first**: the single GPL570 SGK1 probe sits at **8.94 log2 RMA in RS4;11 DMSO - well expressed** - while the 3-prime-end RNA-seq row sits at **1.02 log2 CPM at day 7 and 0.39 at day 2 - at or below the detection floor**. A gene cannot be both. Either the array probe is cross-hybridising, or the 3-prime-end assay's window misses SGK1's dominant 3-prime isoform, and **the second is the more likely** because SGK1 has multiple transcript variants with distinct first exons and the array probe and the sequencing window need not be on the same one [E1]. SEM and REH are both flat on the array (+0.02, +0.03) with high baselines (9.51, 10.89), so the array's RS4;11 +1.00 is at least internally selective. DepMap 26.06: **-0.0417** across 1,258 screens, 0.9% below -0.5 - not a dependency [E2]. PubMed: **0** fusion-scoped records under either SGK1 or SGK, and **0** menin-drug-scoped [E3] - and `SGK[tiab]` returns **658** records sharing **318** with SGK1's 1,818 (Jaccard **0.1474**), the collision the dispatch brief predicted [E3].
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.
The platform disagreement, stated before the result
| dataset | cell | arm | DMSO | log2FC |
|---|---|---|---|---|
| array 201739_at | RS4;11 | MI-2-2 | 8.94 | +1.00 (p 0.043) |
| array 201739_at | SEM | MI-2-2 | 9.51 | +0.02 |
| array 201739_at | REH | MI-2-2 | 10.89 | +0.03 |
| RNA-seq | RS4;11 d7 | VTP-50469 | 1.02 (floor) | +1.99 |
| RNA-seq | RS4;11 d7 | EPZ5676 | 1.02 | +0.07 |
| RNA-seq | RS4;11 d2 | VTP-50469 | 0.39 (below floor) | +0.97 |
| RNA-seq | MOLM13 d7 | VTP-50469 | 2.08 | +0.96 |
An array baseline of 8.94-10.89 log2 RMA and an RNA-seq baseline of 0.39-2.08 log2 CPM are not compatible. On the array SGK1 is among the better-expressed genes in this panel; in the RNA-seq it is one of the worst. Three candidate explanations, in order of plausibility:
- 3-prime-end window mismatch. SGK1 has several transcript variants; a 3-prime-end assay quantifies one window near a polyadenylation site, and if the dominant leukaemic isoform uses a different 3-prime end the assay under-counts it. This is the most likely explanation and it is unfalsifiable from these data.
- Array cross-hybridisation. A single probe with no second probe to corroborate it is exactly the situation where cross-hybridisation is undetectable - and GPL570 carries only one SGK1 probe, so the multi-probe check that saved ATF5 and SESN2 is unavailable.
- Genuine biology: a large difference in SGK1 expression between the batches. Unlikely across the same cell line.
What this does to the result. The +1.99 rests on counts near zero and the +1.00 rests on a single uncorroborated probe. Both point the same way and both are weak, so the direction is worth recording and the magnitude is not. The menin-selectivity (+1.99 versus +0.07) rests entirely on the low-count arm. This entry's confidence is graded accordingly.
Why a druggable kinase going UP is worth an entry anyway
SGK1 is an AGC-family kinase with a canonical ATP-binding kinase domain, four PDB entries, and an extensive medicinal-chemistry literature - it is a tractable small-molecule target in a way that no other member of this signature is. CHAC1, TRIB3 (a pseudokinase with no activity), STC2, SESN2, ATF3 and ATF5 are not druggable objects; SGK1 is.
So the question this entry poses is not "what does SGK1 do" but "is the cell turning on a survival kinase in response to the drug". SGK1 is functionally adjacent to AKT - same family, overlapping substrates, overlapping role in survival signalling - and stress-induced SGK1 upregulation is a documented resistance mechanism to PI3K and AKT inhibition in other cancers. If a menin inhibitor raises SGK1, that is a candidate adaptive-resistance route that no MEN1 sequencing panel would detect, and it is testable: an SGK1 inhibitor plus VTP-50469 in RS4;11, with the prediction that the combination is more than additive.
Two reasons for restraint. (a) The measurement is weak, as above. (b) SGK1's DepMap gene effect is -0.0417 with 0.9% of lines below -0.5, so SGK1 is not a dependency in any unperturbed line - a resistance-mediating kinase need not be a dependency, but it does mean there is no independent functional-genomics support here. The entry is written as a specific, cheap, falsifiable experiment rather than as a finding.
The SGK collision the brief predicted
Rule 88: no SGK1 or SGK row exists in data/symbol-collisions.json (157 rows read 2026-09-22). The dispatch brief predicted this collision and the measurement confirms it 3:
| token | records | shared with SGK1 | Jaccard |
|---|---|---|---|
SGK1[tiab] | 1,818 | - | - |
SGK[tiab] | 658 | 318 | 0.1474 |
318 of 658 SGK records also name SGK1 - so unlike CHOP or CIS2, roughly half the alias's records are genuinely about this gene, and the other half are not (SGK is also used for the family collectively, for SGK2 and SGK3, and in unrelated abbreviations). That intermediate overlap is the awkward case: a 0.15 Jaccard is too low to treat the tokens as interchangeable and too high to dismiss the alias as a different entity. The correct handling is to run both and title-check the non-overlapping set, and for the fusion-scoped question it did not matter: SGK[tiab] AND the eight fusion operands returns 0 with a valid querytranslation and 'No items found.', identical to SGK1's zero.
Contrast with the same beat's PPP1R15A case, where widening to GADD34 turned a zero into the only direct KMT2A-fusion interaction in the panel. Two genes, same beat, same procedure, opposite outcomes - which is why the generalisable rule is "run the historical name and record what it returned", not "the alias is better".
| uniprot | O00141 |
|---|---|
| length AA | 431 |
| entry name | SGK1_HUMAN |
| protein name | Serine/threonine-protein kinase Sgk1 |
| ensembl gene ID | ENSG00000118515 |
| PDB note | UniProt O00141 cross-references four PDB entries and NONE was fetched. No chain, resolution, residue range or ATP-site description is claimed from them. |
| alphafold ID | AF-O00141-F1 |
| accession resolution | gene_exact:SGK1 AND organism_id:9606 AND reviewed:true returns EXACTLY ONE reviewed row, O00141, SGK1_HUMAN, 431 aa, synonym SGK; re-fetched alone with name and length compared back: PASS. |
| localization | FIVE COMPARTMENTS, ALL UNCODED. UniProt O00141 lists Cytoplasm, Nucleus, Endoplasmic reticulum membrane, Cell membrane and Mitochondrion, and there is NO evidence code on any of them - the weakest localisation standing in this beat's panel. UniProt's own note says the subcellular localisation is controlled by the cell cycle and by exposure to specific hormones and environmental stress. Per rule 91 the uncoded statements are recorded and labelled uncoded rather than discarded. No SGK1 localisation experiment exists in any haematopoietic or KMT2A-rearranged cell. |
| subnuclear localization | NOT ASSIGNED and NEVER LOOKED FOR. UniProt lists Nucleus among five compartments with no evidence code and no subnuclear qualifier. No nuclear body, speckle or chromatin address has been reported for SGK1 in any cell type. Recorded as a bounded absence resting on an uncoded prior; no compartment is invented. |
| cell type | RS4;11 (t(4;11) KMT2A::AFF1 - the only responsive line on either platform), SEM (t(4;11), FLAT at +0.02 despite a high array baseline of 9.51), REH (KMT2A wild type, FLAT at +0.03, baseline 10.89) and MOLM13 (KMT2A::MLLT3 AML, +0.96) [E1]. UniProt records expression in most tissues with highest levels in pancreas, then placenta, kidney and lung; isoform 2 strongly in brain and pancreas. No primary leukaemia sample, no normal haematopoietic cell. NOTE: SEM being flat while REH is flat means the array result is selective for RS4;11 and not for the fusion. |
| cell state | Measured in the day-6/day-7 drug-arrested state, and SGK1 is one of the panel members where the state most undermines the measurement: UniProt states explicitly that SGK1's subcellular localisation is controlled BY THE CELL CYCLE and by hormone and environmental-stress exposure, and a drug that arrests the cell cycle changes exactly that. Neither dataset staged or synchronised the cells, and neither recorded serum, hormone or glucocorticoid content - SGK1 is a glucocorticoid-inducible kinase (the name is serum/glucocorticoid-regulated kinase), and glucocorticoids are a backbone drug in this disease. |
| epigenetic readout | No writer, eraser or reader activity. THE EPIGENETIC DATUM IS THE DRUG DISCRIMINATION: SGK1 rises 1.99 log2 under a menin inhibitor and 0.07 under a DOT1L inhibitor, the largest such gap of any UP gene here, which places it with the menin-dependent arm (MEF2C, MEIS1, SOCS2) rather than the H3K79-methylation-coupled ISR arm - but it rests on a low-count arm and should not be weighted like the MEF2C result. No histone-mark measurement at the SGK1 locus exists in any KMT2A-rearranged cell. |
| shown for fusion | MEASURED, WEAKLY AND ONLY IN ONE LINE. RS4;11 (KMT2A::AFF1) rises on both platforms; SEM (also KMT2A::AFF1) is FLAT at +0.02 with a high baseline, and KMT2A-wild-type REH is flat at +0.03. So the array result is RS4;11-selective rather than fusion-selective, and MOLM13 (KMT2A::MLLT3) gives +0.96 on a low-count arm. Transcript only, one array probe, irreconcilable platform baselines. No SGK1 protein, no kinase-activity assay, no ChIP and no functional test in any fusion context. |
| absence class | literature-absence-bounded |
| absence class note | RECLASSIFIED 2026-09-23 (cycle 76). This field held free text, which `scripts/check-absence-class.py` rejects; the prose is preserved here VERBATIM and unsummarised because it is more informative than the token that replaced it. Token chosen: literature-absence-bounded - the two-part prose says MEASURED-AND-<something> for the data arm and BOUNDED-UNPUBLISHED for the prose arm. The data arm is a RESULT, not an absence; the absence is the bounded literature, and the bound's own limits (titles and abstracts only, per rule 80) are stated in the preserved prose. ORIGINAL absence_class, verbatim: «MEASURED-BUT-PLATFORM-DISCORDANT in transcript, direction only; BOUNDED-UNPUBLISHED in prose (0 fusion-scoped under both SGK1 and SGK, 0 menin-drug-scoped, all with valid querytranslations and 'No items found.'), with the limit that the PI3K/AKT-resistance literature that makes SGK1 interesting was not surveyed.» |
PDB IDS
[ "2R5T", "3HDM", "3HDN", "7PUE" ]
domains
[
{
"name": "Protein kinase",
"start": 98,
"end": 355
},
{
"name": "AGC-kinase C-terminal",
"start": 356,
"end": 431
}
]the measurement
{
"rs411_array_mi22_lfc": 1.001,
"rs411_array_mi22_welch_p": 0.0434,
"rs411_array_dmso_log2_rma": 8.94,
"sem_array_mi22_lfc": 0.021,
"sem_array_dmso_log2_rma": 9.51,
"reh_array_mi22_lfc": 0.032,
"reh_array_dmso_log2_rma": 10.89,
"rs411_rnaseq_d7_vtp_lfc": 1.993,
"rs411_rnaseq_d7_epz5676_lfc": 0.07,
"rs411_rnaseq_d7_dmso_log2cpm": 1.02,
"rs411_rnaseq_d2_dmso_log2cpm": 0.39,
"molm13_rnaseq_d7_vtp_lfc": 0.958,
"menin_specificity": "MENIN-INHIBITOR-SELECTIVE - VTP +1.99, EPZ5676 +0.07, difference 1.92 log2, the largest menin-versus-DOT1L gap of any UP gene in the panel. It rests entirely on the low-count RNA-seq arm.",
"THE_PLATFORM_DISAGREEMENT": "array DMSO 8.94-10.89 log2 RMA (well expressed) versus RNA-seq DMSO 0.39-1.02 log2 CPM (at or below floor). IRRECONCILABLE. Most likely a 3-prime-end window mismatch against an SGK1 transcript variant; array cross-hybridisation cannot be excluded because GPL570 carries only ONE SGK1 probe. Direction is recordable; magnitude is not."
}essentiality
{
"depmap_release": "26.06 (read back from meta{dataVersion})",
"depmap_screens": 1258,
"chronos_mean": -0.0417,
"pct_lines_below_minus_0_5": 0.9,
"leukaemia_screens": 57,
"leukaemia_mean_gene_effect": -0.0148,
"reading": "not a dependency in any unperturbed line - which is compatible with a resistance-mediating kinase but supplies no independent functional support"
}identifier traps
{
"registry_status": "NOT in data/symbol-collisions.json as of 2026-09-22 (157 rows read). The dispatch brief PREDICTED this collision and the measurement CONFIRMS it.",
"SGK": "658 PubMed records against SGK1's 1,818, sharing 318 (Jaccard 0.1474). THE AWKWARD INTERMEDIATE CASE: roughly half the alias's records are genuinely about SGK1 and half are not (family-collective usage, SGK2, SGK3, unrelated abbreviations), so the tokens are neither interchangeable nor separate entities. Fusion-scoped SGK returns a valid 0, identical to SGK1's.",
"contrast_with_ppp1r15a_in_the_same_beat": "Widening PPP1R15A to GADD34 turned a zero into the only direct KMT2A-fusion interaction in the panel; widening SGK1 to SGK changed nothing; widening DDIT3 to CHOP added twelve off-topic records. Same beat, same procedure, three different outcomes - so the rule is 'run the historical name and RECORD what it returned', not 'the alias is better'.",
"array": "GPL570 carries ONE SGK1 probe, so the multi-probe cross-hybridisation check that corroborated ATF5 and SESN2 is UNAVAILABLE for this gene.",
"the_platform_trap": "An array baseline of 8.94 and an RNA-seq baseline of 1.02 for the same gene in the same cell line. A cross-platform fold change agreeing in DIRECTION can conceal a total disagreement about whether the gene is expressed at all."
}- SGK1 is functionally adjacent to AKT and stress-induced SGK1 upregulation is a documented resistance route to PI3K/AKT-pathway inhibition in other cancers. A menin inhibitor that raises SGK1 may be selecting for its own escape through a kinase no MEN1 sequencing panel would see - and SGK1 is the only druggable object in this signature, so the countermeasure exists if the effect is real.
- SGK1 is glucocorticoid-inducible by definition, and dexamethasone and prednisone are backbone agents in KMT2A-rearranged ALL. A patient on both a steroid and a menin inhibitor may have SGK1 driven from two directions at once, and no experiment has combined them.
- This entry's measurement is the weakest in the beat and the most likely to be over-quoted, because '+1.99 menin-selective' reads well and the baseline it rests on is 1.02 log2 CPM. Anyone citing the number must carry the platform disagreement with it.
- Is SGK1 expressed in RS4;11 or not? The array says 8.94 log2 RMA and the 3-prime-end RNA-seq says 1.02 log2 CPM. A qPCR with primers in a shared exon, or an SGK1 western, resolves it - and until it is resolved neither fold change in this entry can be interpreted.
- Does an SGK1 inhibitor plus VTP-50469 kill RS4;11 more than either alone? If the drug turns on a survival kinase, that is an adaptive-resistance route that no MEN1 sequencing panel would detect, and SGK1 is the only tractable small-molecule target in this whole signature.
- Why does SEM not respond when its array baseline is HIGHER than RS4;11's? A flat SEM plus a flat REH makes the array result RS4;11-selective rather than fusion-selective, which is the third gene in this beat where SEM breaks the fusion story (CHAC1, ATF5, SOCS2 were the others).
- Does glucocorticoid exposure change this? SGK1 is the serum/glucocorticoid-regulated kinase, dexamethasone and prednisone are backbone drugs in KMT2A-rearranged ALL, and neither public experiment recorded the steroid content of the medium.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
SGK1 rises 1.00 log2 on the single GPL570 probe under MI-2-2 in RS4;11 (p 0.043, DMSO 8.94) while SEM and REH are flat (+0.02 at 9.51, +0.03 at 10.89), and rises 1.99 log2 under VTP-50469 in RS4;11 at day 7 against 0.07 under EPZ5676 - the largest menin-versus-DOT1L gap of any up gene in the panel; but the RNA-seq DMSO baseline is 1.02 log2 CPM at day 7 and 0.39 at day 2, at or below the detection floor, irreconcilable with the array's 8.94-10.89 log2 RMA.
What this evidence cannot show
GRADED MODERATE, NOT ESTABLISHED, BECAUSE OF THE PLATFORM DISAGREEMENT - and that is the entry's first-order fact, not a footnote. An array baseline of 8.94 and an RNA-seq baseline of 1.02 for the same gene in the same cell line cannot both be right. Most likely a 3-prime-end window mismatch against an SGK1 transcript variant, which is unfalsifiable from these data; array cross-hybridisation is also undetectable here because GPL570 carries only ONE SGK1 probe, so the multi-probe corroboration available for ATF5 and SESN2 is not available. The +1.99 rests on counts near zero and the +1.00 on a single uncorroborated probe. Direction is worth recording; magnitude is not. RS4;11 array DMSO arm n=2; array p-value 0.043 is marginal.
Evidence that argues against this (1)
- If the RNA-seq baseline of 1.02 log2 CPM is the truth, SGK1 is functionally absent in RS4;11 and a rise to 3.0 log2 CPM is a rise to still-negligible levels - in which case the 'survival kinase turned on by the drug' reading is empty. If the array baseline of 8.94 is the truth, then the RNA-seq fold change is being computed on a window that is not measuring the gene and the +1.99 is meaningless. Either way one of the two numbers this entry reports is uninterpretable, and nothing here says which. NCBI GEO, re-analysed here 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
UniProt O00141 is the single reviewed human SGK1 record, 431 aa with a protein-kinase domain at 98-355 and an AGC-kinase C-terminal region at 356-431, synonym SGK, four PDB entries, annotated as regulating a wide variety of ion channels, membrane transporters, cellular enzymes, transcription factors, neuronal excitability, cell growth, proliferation, survival, migration and apoptosis with an important role in cellular stress response; its subcellular localisation is stated to be controlled by the cell cycle and by exposure to specific hormones and environmental stress, across cytoplasm, nucleus, endoplasmic reticulum membrane, cell membrane and mitochondrion - and those location statements carry NO evidence codes at all. DepMap 26.06 mean gene effect -0.0417 across 1,258 screens, 0.9% of lines below -0.5, leukaemia mean -0.0148.
UniProtKB - O00141, SGK1_HUMAN, serine/threonine-protein kinase Sgk1 · UniProt Knowledgebase 2026What this evidence cannot show
Resolved live 2026-09-22: exactly one reviewed record (rule 62 satisfied), re-fetched alone with name, length and primary symbol compared back: PASS. RULE 91 FLAG: the entire subcellular-location block for SGK1 is UNCODED - five compartments and no ECO code on any of them - which per rule 91 is recorded rather than discarded but is the weakest localisation standing possible. The four PDB entries were NOT fetched and nothing is claimed from them. The DepMap figures are from the same session with the release read back (rule 83).
Evidence that argues against this (1)
- A kinase annotated in five compartments with no evidence code on any of them, whose distribution is explicitly said to depend on cell cycle, hormones and stress, is not a protein about which a total-transcript measurement in a drug-arrested culture says anything useful. The localisation record actively undercuts the interpretation of this beat's own measurement. UniProt Knowledgebase 2026
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
PubMed returns zero records for SGK1 jointly with an eight-operand KMT2A-fusion list and zero for SGK1 with a thirteen-name menin-inhibitor list, both with non-empty querytranslations and warninglist 'No items found.'; the alias SGK reaches 658 records of which 318 also mention SGK1 out of SGK1's 1,818 (Jaccard 0.1474), and fusion-scoped SGK also returns a valid zero.
NCBI PubMed esearch, run 2026-09-22, with querytranslation and warninglist read back from the response for every query quoted · NCBI PubMed 2026What this evidence cannot show
Rule 3 satisfied on all three zeros. THE DISPATCH BRIEF PREDICTED THIS COLLISION and the measurement confirms it, at an awkward intermediate overlap: 318 of 658 alias records are genuinely about SGK1, so the tokens are neither interchangeable nor separate entities and the alias's non-overlapping set was NOT title-checked here. For the fusion-scoped question it did not matter - both return a valid zero. Rule 94 form; rule 95 observed; rule 80: titles and abstracts only.
Evidence that argues against this (1)
- SGK1 has 1,818 records and a substantial literature as a stress-induced survival kinase and as a mediator of resistance to PI3K/AKT-pathway inhibition in other cancers, none of which is reachable by a KMT2A-scoped query. The zero says the KMT2A field has not looked at SGK1; it says nothing about whether SGK1 would matter if it did, and the resistance literature that makes this gene interesting was deliberately not surveyed by this beat. NCBI PubMed 2026
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
+1.00 log2 on one array probe in RS4;11 and +1.99 under VTP-50469 versus +0.07 under EPZ5676 - menin-selective in direction, but SEM is flat at a higher baseline and the two platforms disagree about whether the gene is expressed
class-level adaptive-resistance candidate: a menin inhibitor raises a druggable survival kinase 1.99 log2 while a DOT1L inhibitor does not. Measured for VTP-50469, not revumenib
SGK1 is the serum/glucocorticoid-regulated kinase and dexamethasone is a backbone agent in KMT2A-rearranged ALL; neither public menin-inhibitor experiment recorded the steroid content of the medium