AGK notes

The PR #3195 follow-up supersedes the earlier non-core plasma-membrane conclusion below. The historical fraction signal is retained as evidence, but its functional compartment attribution is now UNDECIDED.

2026-06-03 - Proteostasis PN review

Falcon deep research findings (2026-06-07)

A Falcon deep-research run completed for AGK (the earlier 2026-06-03 attempt had timed out). The report largely CONFIRMS the existing review; below is a synthesis of KEY findings, flagged CONFIRMS / NEW / PROVISIONAL with provenance. PMIDs resolved via PubMed (DOI->PMID) where possible; Jiang 2020 (Blood) and Vukotic 2017 (Mol Cell) PMIDs could not be reliably resolved in this run and are cited by DOI only (not used to alter annotations).

Evidence re-review, 2026-09-20

Reviewed all 44 rows using lipid enzymology, TIM22 primary studies, source references and PAINT PTHR12358. The ceramide-kinase ancestor PTN008994514 is supported by mouse Agk and fly Mulk, not a CERK-only transfer. Human negative substrate assays (PMID:15939762) remain a target-specific basis for the activity challenge; PMID:16269826 reports a negative MuLK follow-up but also cautions that no firm substrate conclusion could be drawn. The existing OpenScientist report was reused and its incorrect CERK-transfer explanation was explicitly rejected. Ceramide and sphingosine biosynthesis are separately UNDECIDED because lack of phosphorylation does not directly test lipid formation. Plasma-membrane localization was restored as non-core: the full primary fractionation text reports a smaller P4 pool, not absence. Reactome cytosol rows were verified against R-HSA-6802510, the modified AGK(1-33)/BRAF(328-766) fusion participant in all five events; these are not evidence for intact AGK localization. The TIM22 and acylglycerol-kinase core functions remain supported.

2026-09-27 — Complete source and annotation re-review

This section supersedes earlier statements in these notes and the generated PN snapshot that described exclusive mitochondrial localization, removed all plasma-membrane evidence, equated sphingolipid phosphorylation with sphingolipid biosynthesis, or represented the earlier failed research attempt as the only provider outcome. The machine reports and generated PN notes are preserved unchanged.

Identity and baseline: human AGK, UniProt Q53H12, HGNC:21869; MULK is a previous symbol and FLJ10842 an alias. ClinGen's HGNC-attributed gene facts were checked at https://search.clinicalgenome.org/kb/genes/AGK. Parent verified current main 795b693f and zero open PRs for both AGK and MULK. The review contains 44 original assertions, two isoforms and no NEW annotations. The current GOA TSV has 45 lines of assertions because its SMIM26 IPI row is duplicated; the 44 existing source objects are preserved without editing their terms, references, evidence or qualifiers. The current-main HTML differs from the older local rendering and is treated only as a derived publication baseline.

Research execution: the genuine June Falcon report and July OpenScientist hypothesis report were read and preserved. A fresh Falcon attempt with Perplexity Lite fallback was launched concurrently with just fetch-gene-pmids human AGK. It used the normal wrapper's run_deep_research function with the output directed to /tmp/aigr-agk-research/AGK to protect the existing machine report. Both attempts failed before provider execution: deep-research-client dependency resolution could not resolve pypi.org (exit 2 each; wrapper exit 1). Logs: /tmp/AGK-fresh-research.log. Publication caching completed with all 13 existing references already cached. No fresh provider artifact was produced or written manually.

Substrate specificity and propagation

Compartments, complex role and binding

Core and reference audit

Live AmiGO definitions were checked after QuickGO requests failed. GO:0047620 has monoacylglycerol chemical participants (EC:2.7.1.94), whereas GO:0004143 is DAG phosphorylation (EC:2.7.1.107). The original single lipid core used the monoacyl term to represent both reactions. Two substrate-specific lipid cores now capture these distinct activities, plus the kinase-independent TIM22 contribution; there is no redundant umbrella core. The existing contributes_to_molecular_function is preserved for TIM22 and does not assign an autonomous transporter to AGK.

No exact Q53H12 entry occurs in the cached gocams/index.tsv; the AGK text search only matched unrelated NAGK. No GO-CAM was invented and no NEW assertion was needed. All 32 original reference id/title pairs were preserved; two primary references were added. Every reference has a bounded manual assessment. UniProt and the OpenScientist report carry DISPUTED flags for specified content, not an allegation that all their information is invalid. The latter report's CERK-transfer story and claim that Bektas did not test ceramide are contradicted by the inspected PAINT data and human Results.

Local evidence flags were checked for all 15 PMID entries. The 13 existing caches include four with extracted full sections (PMID:15939762, PMID:28514442, PMID:33961781, PMID:34800366). The BioPlex 3 extraction is incomplete despite its full-text metadata; its missing pair table is explicitly acknowledged. Other cached records are abstract-only or bibliographic-only and have full_text_unavailable: true; external reading does not change that flag. PMID:37655851 remains background review context: the inaccessible AGK-specific cardiolipin statement from the prior provider-based finding was removed, without disputing the review itself.

Cache gate: all PMIDs across this notes file and the review YAML were counted. PMID:15252046 and PMID:25002142 are missing after their single normal fetch attempts; the other 13 are present. The review must remain a draft until normal source recovery. No cached publication, source TSV/UniProt record, provider output, hypothesis artifact or generated PN file was modified.

Independent review: root read all 44 decisions, three cores and 34 reference assessments and found no biological blocker. Root independently checked the five fusion-event placements, the monoacylglycerol GO definition and the RHEA:43312 citation against the actual negative human C6-ceramide assay. Aars1's bounded outer-membrane consultation is recorded above. History validation and rendering passed; final targeted validation and the exact file manifest are the publication checks.

Final validation passed. GO:0008320 uses the repository validator's label protein transmembrane transporter activity; current live AmiGO displays transmembrane protein transporter activity with the former as a synonym. The identifier and functional scope agree. The remaining warnings concern the two unavailable primary caches and intentional use of primary evidence rather than provider-report supporting quotes.

2026-09-27 — PR #3195 review follow-up

Read the complete external review on published head 3e201f8136d848005c724c463334348aea63a5cd, including all seven substantive and optional items. The three local gene-file blobs exactly matched the reviewed publication artifacts before edits. The review is now DRAFT, consistent with the missing required caches and the schema's warning semantics. All 44 source assertions, all 34 reference identities, both isoforms and all three core functions remain unchanged. One annotation action changes: plasma membrane KEEP_AS_NON_CORE to UNDECIDED. The resulting counts are 26 ACCEPT, 12 REMOVE, four UNDECIDED and two MODIFY, with no NEW annotations.

Fractionation evidence. PMID:15939762 Figure 2 measures both AGK protein and MOG kinase activity in differential-centrifugation fractions; its cached legend explicitly says, “AGK activity was also determined in each subcellular fraction with MOG as substrate.” The Results identify P2 as the strongest AGK protein/activity fraction and report less protein in P4. Thus the study is not merely a protein-depletion experiment. However, the P4-specific quantitative activity and marker-purity interpretation were not independently resolved in this follow-up: the primary PDF text was retrieved, but the figure-image request timed out. The historical PAINT plasma-membrane IBD at PTN000270644 also remains unreconstructed from the current cache. These limits warrant UNDECIDED for a functional plasma-membrane pool, without declaring the signal to be contamination or proving that no pool exists. This supersedes the non-core interpretation in the September 20 and initial September 27 notes. The original qualifier is unchanged.

Specificity depends on what the assays resolve. Retained MODIFY on the two broad kinase MFs: direct human substrate assays resolve MAG-to-LPA and DAG-to-PA chemistry, providing specific functional replacements beyond the InterPro/ortholog source. Retained ACCEPT on broad membrane CC: membrane association is established, but the combined electronic sources do not uniquely resolve each pool's membrane attachment. The separately supported TIM22 inner-membrane/IMS compartment does not convert every broad source into a measured IMM/IMS assignment. Reasons on all three rows now state that evidence-specific distinction. No new annotation is added for the already-seeded specific kinase activities, and membrane association is not confused with an aqueous IMS location.

Outer membrane and source scope. The Reactome outer-membrane quotation was removed from supported_by because it was the claim being assessed, not independent corroboration. A recovered Hung AGK supplementary hit would verify the identification but an IMS-APEX proximity hit alone would still not distinguish the membrane anchor. The outer-membrane decision stays UNDECIDED pending topology-resolving evidence; independent TIM22 evidence continues to support the inner-membrane core without excluding every other pool. PMID:16269826's cached sentence supports mitochondrial context only, not membrane versus organelle interior or a particular leaflet. Its IDA membrane annotation remains accepted with curator deference and independent human membrane evidence; the rationale and reference assessment now state that distinction explicitly.

Core contribution and ontology. Added an expert question explaining that GO:0008320 is a contributes_to_molecular_function synthesis for the TIM22 structural subunit, not a new independent transporter assertion. The complex role, qualifiers and all cores are unchanged. Live AmiGO GO:0008320 still lists “transmembrane protein transporter activity” as its name and “protein transmembrane transporter activity” as a synonym; the existing authored label accepted by the repository validator is retained. No new ontology ID is invented and no NEW process is proposed.

The notes-inclusive cache census remains 15 PMIDs, 13 present. Missing PMID:15252046 and PMID:25002142 remain publication draft gates. Prior genuine provider reports and all machine caches remain untouched; this bounded review follow-up did not retry failed provider/fetch jobs. Targeted gene validation, history validation, source integrity and rendering are checked in the final follow-up receipt. No Git state, remote comments or shared project files were changed.

The coordinator independently read the complete follow-up delta and confirmed source preservation, with no biological blocker. Targeted gene validation passes with the two expected warning categories (missing caches and no provider-report supporting quotes); final schema, history validation and rendering pass.

2026-09-27 normal publication-cache recovery

Both required source records are now present. PMID:15252046 is abstract-only,
so its local full-text limitation remains. Its positive recombinant MuLK
substrate results do not resolve the previously documented human assay conflict
or unread donor biosynthetic-process evidence. PMID:25002142 now includes
XML full text from PMC4743503. The inspected Results and Methods support
IMS-targeted APEX proximity with ratiometric controls for cytosolic background.
Biotinylated peptides in selected known IMM proteins face the IMS. The extracted
main text does not expose the AGK-specific supplementary hit, and recovering
the paper does not by itself resolve AGK outer- versus inner-membrane attachment.
That annotation remains UNDECIDED. Its local full_text_unavailable flag is
now false, with the supplementary evidence limit stated separately.

The exact records originate from normal fetch output in Actions run 36286975328,
head 5946477c8ac79ade0709264c775ea1262b108438, artifact 10920674630.
The verified transported ZIP SHA-256 is
c0ffe4a66b80278af34b44aab6a3ae354ffd5699236b3a486ca95527be5e9713;
tmp/verified-reference-records/local-import-receipt.json records per-file
hashes. Only these two gene-required caches enter this follow-up manifest.
No cache, genuine provider report or other machine artifact was edited.

All 44 source assertions and reviews, three cores, 34 reference identities,
isoforms and prior history remain unchanged. This dated entry supersedes the
missing-cache status above. Targeted validation, rendering and history checks
are recorded in the closure manifest. The intentional provider-quotation
advisory, if still emitted, keeps the YAML DRAFT under the schema's zero-warning
COMPLETE rule; it is separate from the now-closed source-cache gate. No provider
text was promoted into primary evidence merely to suppress that advisory.

2026-09-27 — Post-merge source12 assessment and recursive bibliography audit

The baseline is the merged review from PR #3195. All 25 gene/provider/history files matched imported main 39b086a55856ca2f2287036a5ab43280add1ee90 and fresh main 03db1457249828e300303e015f99c334168f689c before authoring. Human AGK (HGNC:21869, Q53H12), previous symbol MULK and alias FLJ10842 were checked separately; neither alias directory nor an overlapping open gene-review PR was found. The 44 annotation objects and their decisions, all 34 prior reference identities, three core functions and both products are unchanged. One inherited PMID:15939762 finding quote gains the literal parentheses in “(AGK), that” to match its cached abstract; the prior reference assessments and other findings are unchanged. Counts remain 26 ACCEPT, 12 REMOVE, four UNDECIDED and two MODIFY, with no NEW annotation or new core term.

Four DOI-linked papers cited by the genuine Falcon report are now available as exact normal-fetch source12 records. Their PubMed identifier, title and DOI were checked independently. Two provide full scientific sections and two remain abstract-only; cache recovery is not treated as proof that every experimental detail was inspected.

Source Actual access and evidence Interpretation for AGK
PMID:23266196 Abstract: two human Sengers pedigrees with p.M1I or p.K327*, citrate-synthase crystals/activity, tissue-dependent respiratory-chain defects and elevated MnSOD protein. DOI 10.1016/j.ymgme.2012.11.282. Supports human disease and mitochondrial dysfunction. The lipid/redox explanation in the abstract is an interpretation, not an additional direct AGK substrate assay or an experimental separation of kinase and TIM22 functions.
PMID:34948281 Full Results 2.1–2.3, Methods 4.1–4.8 and Conclusions: patient-fibroblast cDNA from homozygous c.518+1G>A is 96 bases shorter, consistent with exon-9 skipping; Seahorse respiration and spectrophotometric complex-I/V defects are measured. DOI 10.3390/ijms222413484. Human transcript and respiratory defects are demonstrated. Section 4.8 uses SWISS-MODEL, so the mutant protein model is not a measured kinase or TIM22 experiment. The Conclusions explicitly leave the relative kinase/import contributions unresolved.
PMID:41695748 Full Case report, genotyping, Figure 3 and Discussion: maternal p.Arg137* plus a paternal 7.6-kb exon-5 deletion in a human infant; sequencing/CNV, breakpoint PCR/Sanger and segregation establish the diagnosis. DOI 10.3389/fped.2026.1714952. The paper explicitly lacks patient-derived functional validation. Its deletion-formation interpretation does not establish a DNA-repair function for AGK, and it does not measure the mutant protein, lipid kinase or import mechanism.
PMID:32202634 Abstract: mouse megakaryocyte/platelet-specific AGK loss, G126E comparison, AGK–JAK2 binding and JAK2/STAT3 signaling. DOI 10.1182/blood.2019003851. Positive contextual regulation is supported: kinase-dead G126E retains platelet-count and megakaryocyte-differentiation outcomes. This neither makes AGK the JAK2 protein kinase nor establishes the same mechanism in normal human physiology. Full Methods were not recovered in this record.

The two full human case papers were also read independently by annotation_aars2, who confirmed the distinction between transcript/respiration measurements, structure prediction and unperformed mutant-function experiments. These sources do not contradict the existing MAG-to-LPA, DAG-to-PA or kinase-independent TIM22 cores. Four bounded reference assessments are added, bringing the reference list to 38; no context-specific disease phenotype is promoted into a NEW function.

Retained reports and the complete source boundary

The census includes the YAML, this notes file, generated PN notes, the genuine Falcon report and artifact, and the OpenScientist hypothesis report, citations file, raw HTML, PDF text and four provenance JSON files. The raw reports remain unchanged, including the already-disputed ceramide-transfer narrative. The input artifacts contain nine distinct DOIs, three PMCs and 32 explicit PMIDs; four DOI-to-PMID mappings add the recovered papers above, producing 36 distinct required PMIDs. The three PMC links resolve to already-counted PMID:25002142, PMID:32901109 and PMID:37009826. The census does not recursively add unused references from the bibliographies of these papers or the immutable UniProt record.

Seventeen substantive citations in the retained hypothesis report were absent locally and on fresh main. Their primary identifier/title mappings were checked using official indexed PubMed/PMC records; five CERK-comparator identities were independently checked by annotation_aars2. For PMID:40536506, the successful primary mapping route was the NCBI PubChem Gene bibliography; the direct PubMed view was blank. Identity verification does not certify the full experiments or endorse the provider's interpretation. The finite remaining PMID set is:

PMID Verified DOI Source role in the retained report
PMID:18004883 10.1021/bi701584v Lipid-substrate enzymology
PMID:19112101 10.1152/ajplung.90431.2008 Human bronchial-cell LPA/EGFR signaling
PMID:22069480 10.1371/journal.pone.0026993 Drosophila Dcerk/Dmulk genetic study
PMID:22864860 10.1007/s00592-012-0422-1 Diabetic-retinopathy expression biomarkers
PMID:31303091 10.1080/15513815.2019.1639089 Human postmortem genetic diagnosis
PMID:34368119 10.3389/fcell.2021.659158 Oncology review
PMID:37051931 10.1161/ATVBAHA.122.318647 Mouse platelet activation/thrombosis
PMID:38556546 10.1038/s12276-024-01203-4 CERK comparator
PMID:39636206 10.1158/1078-0432.CCR-24-1192 CLL AGK/JAK2 signaling
PMID:39824030 10.1016/j.nmd.2024.105271 Human clinical genetics
PMID:40536506 10.1097/hc9.0000000000000731 CERK comparator
PMID:40943293 10.3390/ijms26178374 CERK review
PMID:41226412 10.3390/ijms262110373 CERK comparator
PMID:42026151 10.1038/s42003-026-10073-7 Mouse oocyte AGK loss
PMID:42212382 10.3892/or.2026.9145 Pancreatic-cancer NF-κB signaling
PMID:42285305 10.1016/j.jlr.2026.101077 CERK comparator
PMID:42391447 10.1002/mc.70134 Glioblastoma treatment response

These primary abstracts are not interchangeable with the provider's claims. In particular, PMID:22864860 measures expression biomarkers in human vitreous and diabetic-rat retina; it is not an independent purified-AGK substrate experiment. The platelet-production G126E findings in PMID:32202634 and platelet-activation findings in PMID:37051931 concern different outcomes. Neither abstract alone establishes a direct protein-substrate reaction by AGK. The CERK-comparator papers do not substitute for human AGK assays. No annotation action is changed on the basis of these incompletely inspected sources.

The DOI-only dissertation 10.53846/goediss-10678 is retained as a bibliographic source, without an invented PMID. Indexed primary text from the Göttingen repository PDF identifies Anusha Valpadashi's TIM22 thesis and includes human TIM22 crosslinking-mass-spectrometry context. Direct repository/PDF access returned 403, so the complete dissertation was not read and no new experimental claim is based on it. This is a separate access limitation from the missing normal publication caches.

Seven Reactome records are cited. Six event caches are present; the missing physical entity R-HSA-6802510 was independently rechecked on its official page. It identifies human AGK(1–33)InsW-p-BRAF(328–766) in cytosol, with BRAF sequence inserted after the AGK N terminus. This directly grounds the already-retained fusion-context distinction, without excluding a native cytosolic AGK pool. One ordinary cache_reactome_pathway attempt failed because reactome.org could not resolve; no source record was manufactured.

A separate preflight found that the local PMID:37009826 cache lacked only the published full_text_attempted: true metadata line. Both byte sequences were saved, the coordinator inspected the exact diff, and the canonical file was restored from verified current-main bytes under explicit authorization. All scientific content is identical; this restoration is not a new PR file because the restored blob is already present in the publication base. Other cached sources, provider artifacts and prior histories are unchanged.

The source12 records are bound to tmp/source12-canonical-import-receipt.json; the source assessments and byte hashes are in tmp/AGK-source12-followup/assessed-sources.json. The finite missing-source identity receipt is tmp/AGK-source12-followup/missing-source-identities.json. A single ordinary 17-PMID fetch was launched for the newly discovered set; its terminal result and final cache census are recorded below and in the follow-up manifest. The review remains DRAFT for the unresolved source gates and the intentional provider-quotation advisory.

The ordinary publication fetch finished with exit 1, cached 0/17, with a DNS error for each requested PMID. No candidate file was generated. The final required census is 36 PMIDs (19 present, 17 missing), seven Reactome records (six present, one missing), and 26 DOI identities after adding the explicit missing-source mapping table. The dissertation remains a separately documented DOI-only access limit. Targeted gene validation passes with one intentional provider-quotation advisory; the notes/provider-only missing sources are an independently checked publication gate, not a claim that the validator detected all of them. All 98 ordinary quotations pass case-sensitive whitespace-normalized checks, with UniProt line-format prefixes removed for comparison; none requires case folding. There are no special full-text excerpt fields. History, rendering and the final byte/source-preservation checks pass.

2026-09-27 — recovered hypothesis sources and complete AGK cache closure

This entry supersedes the preceding source12 missing-cache counts. All 17 remaining publication records and the AGK–BRAF physical entity are now present as unchanged normal-fetch source22 files, bound to tmp/source22-canonical-import-receipt.json. The earlier source12 manifest and history remain unchanged. Before authoring, all expected baseline blobs matched fresh main 1b6c93bd934b9df5970a0b84680a26b19d5c0083; canonical and alias searches found no overlapping open review. The publication comparison remains against the merged AGK baseline, with the four unpublished source12 publication records and its history included alongside this follow-up.

All 44 complete annotation objects, three core functions, two products and 38 preceding reference assessments remain unchanged. Counts remain 26 ACCEPT, 12 REMOVE, four UNDECIDED and two MODIFY; no NEW assertion or core term is added. Eighteen source-specific reference assessments distinguish actual experiments from contextual citations. Six new publication records contain abstracts only; eleven contain scientific body sections (including two reviews). The Reactome record is entity metadata only. Relevant Methods, Results and Discussion were read where actually present; this does not claim every linked supplement was recovered.

Source Actual evidence and limits
PMID:18004883 Abstract only. The abstract compares MuLK phosphorylation of sn-1,2- and sn-2,3-dioleoylglycerol, with lower stereoselectivity than the compared DAG kinases. Species and enzyme-preparation details are not explicit in this cache. This is positive lipid-substrate evidence, but not an independent human ceramide-phosphorylation assay.
PMID:19112101 Abstract only. The abstract reports human bronchial epithelial AGK knockdown and lentiviral wild-type overexpression, intracellular LPA changes and altered LPA-induced IL8, EGFR and MAPK responses. Microscopy localizes overexpressed AGK to mitochondria. These cellular results support contextual lipid-signaling regulation; they do not isolate direct AGK phosphorylation of EGFR.
PMID:22069480 Scientific body sections. Relevant Results, Figures 6–8, transgenic-line and lipid-MS Methods were read. Drosophila CG31873/Dmulk coding sequence is expressed from UASp constructs; Dcerk/Dmulk genetics affect primordial-germ-cell migration, and Dmulk overexpression increases whole-embryo C1P approximately 3.3-fold. This is genuine positive fly evidence. The inspected experiments do not isolate purified Dmulk or human AGK ceramide phosphorylation, and cited human/mouse enzymology is background. The XML repeats nested sections; linked supplements were not independently read. This source does not erase the distinct human negative-substrate assays.
PMID:22864860 Abstract only. The abstract describes 42 human proliferative-diabetic-retinopathy vitreous samples, 35 nondiabetic controls and diabetic rat retinas, using ELISA and immunoblotting. Expression and biomarker associations do not establish a new AGK substrate or direct angiogenic mechanism. Full Methods and Results are unavailable locally.
PMID:31303091 Abstract only. The abstract reports postmortem diagnosis in a nine-month-old human infant with homozygous c.1215dupG/p.Phe406Valfs*4. It supplies clinical genotype/phenotype evidence, not a variant-specific kinase, protein-import or purified-protein assay. Full Methods and Results are unavailable locally.
PMID:34368119 Scientific body sections. The local review body was read for lipid-kinase, kinase-independent TIM22, Sengers syndrome, cancer and immune-cell sections. It synthesizes mixed human and mouse studies rather than reporting new experiments. Its protein-kinase heading and broad disease framing do not independently establish a protein substrate for AGK.
PMID:37051931 Abstract only. The structured abstract describes transgenic mouse platelet activation and thrombosis experiments, co-immunoprecipitation/mass spectrometry, staining, immunoblots and PA/LPA measurements. AGK deficiency or mutation reduces activation, while measured PA/LPA output is unchanged; Talin1 Ser425 phosphorylation is implicated. The authors report kinase-dependent effects, but the abstract does not resolve a purified direct AGK-to-Talin1 phosphotransfer reaction. This platelet-activation outcome differs from the kinase-independent platelet-production phenotype in PMID:32202634; neither result is discarded.
PMID:38556546 Scientific body sections. Relevant local Methods, Results and Figures 2–8 concern CERK/C1P in mouse CCl4 and ischemia-reperfusion liver injury, mouse AML12 cells and human primary hepatocytes. Mouse KEAP1/domain-deletion constructs are expressed in HEK293T cells. C1P-coated beads and domain deletions implicate the lipid as the KEAP1 ligand, not the CERK protein or AGK. Docking and lysate pulldown are not a purified binding-constant assay. Supplementary-only methods were not independently recovered. This is a distinct-enzyme comparator, not direct AGK evidence.
PMID:39636206 Scientific body sections. Patient/cell, siRNA, co-immunoprecipitation and microscopy Methods and the AGK/JAK2 Results were read. Primary untreated human CLL samples and normal B cells show cellular AGK/HSP90 and AGK/JAK2 associations, AGK-siRNA effects on phosphorylated JAK2, and cytoplasmic/nuclear AGK localization. JAK2 performs the proposed H3 Tyr41 phosphorylation; AGK is not demonstrated to be a histone kinase. These native-AGK observations are independent of the AGK–BRAF fusion events and do not make those events evidence for native-AGK localization. Supplements were not independently read.
PMID:39824030 Abstract only. The abstract reports an adult Palestinian woman with homozygous c.221+1dup and a Sengers syndrome phenotype. Its TIM22 and respiration discussion is background, not a reported variant-specific import or kinase experiment. Full Methods and Results are unavailable locally.
PMID:40536506 Scientific body sections. Local animal/cell Methods, Results and p38-inhibitor experiments use C57BL/6 Cerk-knockout mice, explicitly murine Cerk adenoviral overexpression and mouse AML12 hepatocytes. CERK perturbation changes p38/HSPB1 phosphorylation, and SB203580 reverses protective cellular effects. This is pathway epistasis, not purified CERK phosphorylation of p38 or HSPB1, and not an AGK experiment. Human transcriptomic associations do not change the experimental protein/model scope.
PMID:40943293 Scientific body sections. The local article is a review of CERK-generated C1P and cell growth/survival signaling, with no new functional assay identified in the inspected body. It is useful comparator context, not independent primary replication or evidence that AGK catalyzes ceramide phosphorylation. The lipid and its generating enzyme remain distinct entities.
PMID:41226412 Scientific body sections. Local Results and Methods 4.2–4.10 measure radiolabeled ATP-dependent C1P formation in enriched pig-kidney basolateral membrane fractions, with TLC and protein/time/boiling controls. Separate human HK-2 lipidomics detects C1P and human CKD datasets provide expression context. The enzyme assay is not purified human CERK and does not identify AGK as its catalyst. cAMP/PKA effects and assay temperature/pH optima do not establish a new human AGK function.
PMID:42026151 Scientific body sections. Conditional-allele/animal Methods and mitochondrial, fertility and metabolomic Results use mouse Agk exon-3 flox crossed with Dppa3-Cre. Oocyte loss and mitochondrial defects establish a physiological requirement; PA rises despite knockout and LPA is undetectable. No direct human AGK or TIM22 import assay is performed. Unchanged TIM22 transcript abundance does not exclude a protein-complex contribution, and kinase-dead knock-in is proposed as future work. The measured requirement does not by itself establish an additional catalytic core.
PMID:42212382 Scientific body sections. Local Methods explicitly identify full-length human AGK in N-terminal Flag-pcDNA3.1, human AsPC1/PANC1 cells and pancreatic tissues; PANC1 xenografts use mouse hosts. Knockdown/overexpression, NF-kappaB reporter, p65 localization/phosphorylation and partial epistasis support cellular pathway regulation. They do not isolate direct AGK phosphorylation of p65 or IKK. Supplementary/raw data were not independently assessed.
PMID:42285305 Scientific body sections. Local Methods and Results use HeLa CERK knockout/siRNA/inhibitor models, HA-tagged wild-type versus kinase-dead G198D rescue, and separate mouse Cerk-knockout cerebellar proteomics. Wild type restores NBD-C1P production and LC3B-II where G198D does not. The inspected Methods do not establish the rescue-insert species accession from the human host alone. The GFP/RFP reporter lacked a significant starvation difference, and the authors limit conclusions to LC3B-associated autophagosome formation rather than global autophagy. These are CERK experiments, not AGK evidence.
PMID:42391447 Scientific body sections. Local cell/construct Methods and Results 3.1–3.6 use human GBM lines U251/LN299 (as printed), shRNAs, a full-length AGK cDNA without an explicit species accession, and 82 human GBM cases. Quantitative Results show reduced basal proliferation but increased temozolomide resistance and senescence after AGK depletion. Some concluding wording reverses that direction; the assessment follows the actual Results and corresponding Discussion. The AGK-to-ROS mechanism remains unresolved, and no purified lipid- or protein-substrate assay is presented. Human host is not treated as proof of an unspecified insert origin.
R-HSA-6802510 The 228-byte local record contains only the stable identifier, display name and primary link. The official page independently establishes the phosphorylated AGK(1–33)InsW-BRAF(328–766) fusion. This is the specific fusion source behind the five cytosolic event assertions, not a native-AGK localization assay.

The new human CLL observations permit a native cytoplasmic/nuclear pool in that cellular context; they do not repair the attribution of five different fusion-specific Reactome assertions. The established mitochondrial catalytic and TIM22 cores remain supported. Similarly, positive fly Dmulk/C1P data are retained as donor evidence without replacing the human substrate-assay findings. The CERK comparator studies do not assay AGK. Their lipid products, downstream kinases and cellular phenotypes are not additional AGK catalytic activities.

The complete recursive census covers authored YAML/notes, PN notes, genuine Falcon and OpenScientist reports, their citations, HTML, PDF extraction and provenance JSON, plus the regenerated review HTML. It resolves 36 required PMIDs, seven Reactome records and 26 distinct DOI identities. All required normal PMID and Reactome records are now present. The DOI-only Göttingen dissertation 10.53846/goediss-10678 remains separately identified from primary repository metadata; the complete body was not recovered, no PMID was invented, and no new experimental assertion depends on it. Recovery closes the cache gates, not the documented full-text or assay-specific uncertainties.

Independent source identity checks are recorded in tmp/source22-primary-identity-review.json; actual AGK article/abstract scopes are in tmp/source22-agk-scientific-review.json. The five CERK comparator bodies were independently read in tmp/source22-cerk-scientific-review.json, including explicit model, construct and pathway limits. Raw sources, genuine provider artifacts and published histories are unchanged. The YAML remains DRAFT for its intentional provider-quotation advisory; that advisory is distinct from the now-closed source-cache gate and does not itself require a draft PR.

Targeted validation passes with the single intentional unused-provider quotation advisory. All 98 existing quotations pass case-sensitive, whitespace-normalized source checks; history and rendering pass, with no doubled PubMed links or new YAML aliases. The independent bounded peer read all 18 added assessments and the appended notes, independently confirmed the 44/3/2 and preceding-reference preservation, and rechecked the MuLK abstract, mouse platelet abstract, oocyte study and GBM treatment-response sections. No material issue was found; the peer did not claim a second complete read of every paper or independently certify the recursive census.

PR review follow-up, 2026-09-27

Review of PR #3314
requested that recovered ortholog evidence be linked directly to the ceramide-kinase decision.
PMID:22069480 and PMID:18004883 are now cited on that annotation. The PAINT source comment
names the positive Dmulk result (approximately 3.3-fold increased embryonic C1P on overexpression),
with an exact cached excerpt. This in-vivo fly lipid result does not isolate human ceramide
phosphorylation. The human substrate evidence and REMOVE decision remain unchanged; all 44
annotation actions, original source objects, two products and three core functions are preserved.

A new research question connects mouse platelet PA/LPA measurements PMID:37051931 and mouse
oocyte metabolomics PMID:42026151 to direct catalytic capability, steady-state lipid abundance
and secondary mitochondrial effects. These findings do not invalidate the human in-vitro MAG/DAG
activities or establish a new reaction. The fly study is rated MEDIUM for supporting ortholog
context. Reference assessments use ordinary source-access wording instead of session batch names.
The five CERK comparator papers remain LOW because they document a specific attribution problem
in the research summary; that purpose is now explicit. Previously recorded local tmp paths are
operational journal details, not required scientific sources. The publication caches, annotation
citations and PR provide the reviewable evidence trail.