HGNC:362 is the approved human symbol AK2; the archived official HGNC subset
has no previous or alias symbols. UniProt P54819 also lists ADK2. Fresh direct
GitHub API checks at main 30a9290881824baaacab2b681d7808f572c4cef6
found the canonical five gene paths and task branch absent, with no open
canonical-title PR. Separate ADK2 path/title checks found no alias directory or
open PR. All three exact source3 seed hashes match the import receipt; no
reviewer-authored baseline existed. The 30 seeded assertions, 14 original
references and six alternative products are preserved for audit. Publication
base d35dcc30b44924f79c0510b281ae824aa536848a also lacks the gene paths.
Default Falcon research with --fallback perplexity-lite --timeout 1200 and
normal fetch-gene-pmids were launched concurrently. Both provider commands
failed before contacting a provider because uvx could not resolve PyPI to
install deep-research-client; each returned 2, and the wrapper returned 1.
No provider report was produced. The cache command returned 0 and found all
five seeded PMID records already cached; it did not refresh or modify them.
Logs: /tmp/AK2-provider.log and /tmp/AK2-seed-fetch.log. Research proceeds
through cached primary sources and separately documented primary web access.
The 30 original assertions were reviewed individually. Their source objects,
qualifiers, supporting entities and six alternative products remain unchanged.
Eight generic protein-binding rows are removed as uninformative; this does not
reject their reported interactions. Three broad electronic function/process
rows are refined to measured AMP kinase activity or adenylate interconversion.
Two sperm locations are retained as contextual ortholog inferences. Exosomal
localization remains unresolved because its target-specific supplement was not
recovered. One new molecular-function proposal is described below.
PMID:6182143 was verified against PubMed, and its original JBC full article was
read through the author-uploaded text on ResearchGate publication 17011438
on 2026-09-27 (DOI 10.1016/S0021-9258(18)33631-7). Methods identify human
liver and separate coupled assays for both directions; Results and Table III
support adenine-nucleotide phosphate transfer by purified mitochondrial enzyme.
The local cache is metadata-only and was not modified. PMID:9504408 independently
reports active purified recombinant human AK2A and AK2B, not all six products.
The exact Reactome:R-HSA-110144 and Reactome:R-HSA-110145 summaries separate
human enzymology from rat-derived intermembrane-space inference. UniProt agrees
with that compartment. The human MitoCoP study (PMID:34800366, cached full text)
is retained at its original broad mitochondrion resolution, with curator
deference because the AK2-specific supplement was not re-extracted. It does not
supply an independent intermembrane-space assay. GO:0005737 is cytoplasm,
including cytoplasmic organelles, rather than cytosol; the inherited broad
location is therefore appropriate. Apoptotic release reported in PMID:10218571
is a contextual mammalian observation, not the basis for constitutive cytosol
as a core location.
Live AmiGO definitions were checked for GO:0004017, GO:0005737, GO:0006172,
GO:0015949, GO:0055086 and GO:0072542. ADP formation is direct chemical work by
AK2, not synthesis of the adenine/ribose backbone. AMP and ATP metabolism are
also direct substrate/product participation. The one integrated metabolic core
retains those scopes without manufacturing a differentiation or ATP-hydrolysis
annotation.
The cached interpro/panther/PTHR23359/paint.tsv includes the exact IBD assertions
at PTN000599576 (kinase and cytoplasm), PTN000599693 (ADP biosynthesis), and
PTN001916109 (mitochondrion). Only the PTNs are entered as IBA source entities.
Human AK2 among the kinase descendants is valid experimental grounding, not
circular support. The family tree/MSA and all loss events were not reconstructed.
InterPro signature membership was checked against the immutable UniProt record;
RHEA:12973 and EC:2.7.4.3 match the direct reaction. ARBA/UniRule condition
internals remain unresolved independently of the supported target functions.
MGI's comparative Ak2 graph
(snapshot 2023-03-10) traces both sperm-compartment donor terms to mouse IDA
PMID:16790685. The primary abstract specifically identifies AK2 in the sperm
midpiece mitochondrial sheath, distinct from AK1 elsewhere in the flagellum.
Mouse Q9WTP6/ENSMUSP00000030583 is the donor; no direct human sperm assay was
claimed. Conserved mitochondrial targeting makes a contextual transfer
plausible. Full source was not recovered, so an axonemal structural or motility
mechanism is not inferred.
PMID:20458337 has an abstract-only cache. The original author-uploaded main
article
was inspected, but its AK2-specific supplementary peptide row was not recovered.
Its total exosome proteome and MHC-II co-immunoprecipitated subset are different
assays. The annotation remains UNDECIDED; no claim of contamination or false
localization follows from the dominant mitochondrial pool.
The original PMID:32814053 institutional full PDF
was read. It describes repeated human Y2H screens and integration with other
interaction evidence. The eight exact partners are also listed in the immutable
UniProt record. No AK2-specific functional mechanism for those partners was
established from the main paper, and pair-level supplements were not reanalyzed.
REMOVE implements generic-binding policy, not a finding that these pairs do
not bind. DUSP26 is not substituted for one of these unrelated pairs.
PMID:24548998 was independently verified at PubMed,
with full primary Methods/Results read through the Nature PDF
and indexed PMC3948464 (DOI 10.1038/ncomms4351). Figures 3–4 directly test
recombinant human AK2 binding and activation of DUSP26. The phosphatase acts on
pNPP and Plk-phosphorylated FADD; kinase-defective AK2 K28E retains activation.
AK2 alone has no measured phosphatase activity, and AK3/FADD controls do not
stimulate DUSP26. Tagged proteins and gel-filtration fractions do not establish
fixed endogenous complex stoichiometry. Independent annotation_a4galt reading
confirmed these assay boundaries and found no blocker for the activator MF.
NEW checks: AK2 performs activation, DUSP26 performs dephosphorylation.
GO:0072542 requires binding and increased protein-phosphatase activity; its
parents GO:0019211/GO:0019888 are regulator terms. Its narrower tyrosine-specific
child is not appropriate for the FADD phosphoserine assay. Neither this term
nor its ancestors/descendants occur in the seeded AK2 set. Local AMBRA1 and
CALM1 GOA files contain experimental GO:0072542 annotations, establishing
comparable noncatalytic phosphatase-activator usage. These database rows are a
term-usage check, not substituted experimental evidence for AK2. An exact
P54819/AK2 search of gocams/index.tsv found no target activity. The proposal
is kept distinct from the dominant metabolic core because its general tissue
importance is unresolved. No new cell-proliferation, apoptosis or differentiation
process is proposed. PMID:17952061's primary abstract is retained as signaling
context, with full construct/control details unresolved.
The cached PMID:19043417 human genetics/zebrafish abstract, primary
PMID:19043416 abstract and Figure 3,
and PMID:39378586 with indexed full
PMC11830988 establish disease context. Human patient rescue and progenitor
perturbations link nucleotide imbalance and stage-specific metabolic control to
reticular dysgenesis. They establish AK2 necessity; no direct developmental
process is manufactured solely from loss/rescue phenotypes. The latter paper's
DOI is 10.1182/blood.2024024123. Human disease and zebrafish experiments are not
interchanged.
One normal six-ID cache request completed with exit 1 and 0/6 records recovered:
PMID:16790685, PMID:19043416, PMID:39378586, PMID:17952061, PMID:10218571,
PMID:24548998. Every failure was DNS resolution; log /tmp/AK2-extra-fetch.log.
External primary access does not create a synthetic cache or change local
full_text_unavailable flags. These six remain publication gates. The genuine
provider attempts produced no artifact, so there are no provider-only sources.
The final census includes authored YAML/notes and their decoded DOI/URL
citations, while immutable raw UniProt bibliography is distinguished from
reviewer-cited evidence. No raw data, publication, Reactome or provider bytes
were edited.
The coordinator independently read all 31 decision objects, propagation
assessments, 22 reference assessments, the integrated core and questions, and
checked the primary DUSP26 activation controls and live term definition.
Biological review passed. A requested Reactome quote-boundary cleanup was
applied without changing its meaning or any decision.
Final just validate human AK2 passed with one grouped warning covering the
six missing records above; status remains DRAFT. An earlier run found the
singular failure_mode spelling, corrected to the schema's failure_modes
list before the final full rerun. History validation and rendering passed.
Independent preservation checks confirmed all original source objects,
reference IDs/titles and alternative products; 35 cached quote occurrences
match their immutable sources. The one new DUSP26 excerpt was verified against
the primary PubMed abstract but cannot be checked against a local cache yet.
The recursive authored citation census is 12 PMIDs and three Reactome records;
all DOI occurrences map to those same records, and no provider artifact exists.
Exact hashes, existing-base reuse and five missing-base immutable source3
publication/Reactome dependencies are listed in the frozen manifest.
The current draft PR head was independently checked as
df920d43fa4a890d1ddaec0510c7aef78b1ab0b1; all five canonical gene files
matched its exact blobs before editing. The full formal review 5329953434 and
detailed comment 5855075916 accept the biological decisions and identify the
six missing records as the only blocking issue. The coordinator imported the
normal source14 records verbatim, with receipt
tmp/source14-canonical-import-receipt.json (SHA256
45e4e452e6eaad0c359bc0ff3d8ba2006bde464894cd5dd3cc2af5c073a5b74e).
The source run is 36308419697 at commit
4d08457ceb0e3ce1277c73f918c0746b65866ab5; the artifact SHA256 is
aaefb946ed96a34eda5f323f7b71d8c7fb1c6e233d3b1f1ecaf46c65cc5b94d9.
The earlier missing-cache statements above describe the original session and
are superseded by this dated recovery record, without rewriting its history.
All six actual records were read with their available scope:
The seven optional review suggestions were assessed. The correct PAINT file
is interpro/panther/PTHR23359/PTHR23359-paint.tsv, superseding the shortened
historical filename above. Its actual GO:0004017 IRD records occur at
PTN000599579, PTN002764060 and PTN002764136. The TSV has assertions and taxon
labels, but no full tree or alignment: no independent target-lineage topology
reconstruction is claimed from those labels. Current IBA prose emphasizes
positive human evidence and the inspected IBDs, while this note records the
bounded IRD check. The previous named internal peer identifier denotes an
independent peer review and is not a biological source.
The description now states AK2's biology directly; isoform and experimental
scope remain in the annotation reasons, references and questions. The existing
GO:0055086 refinement already explicitly says that the specific interconversion
term is represented by the Reactome row; source-level duplicates are allowed,
and no additional NEW term is being proposed. Broad cytoplasm remains correct
at the source's resolution. All eight generic-binding removals now also quote
the cached PMID:32814053 description of the systematic human interaction
network, alongside their distinct UniProt pair evidence. That network-level
excerpt is not represented as a reanalysis of individual supplementary pairs.
All 30 seeded assertions, the one prior authored NEW, every action, six
alternative products, and the integrated catalytic core remain unchanged.
The recovered records keep their exact normal-fetched bytes and titles.
Validation, the final recursive decoded PMID/DOI/PMC/Reactome census and the
explicit publication manifest are completed separately after this source read.
Final checks passed: all 48 supporting-text occurrences match their immutable
local sources; all 31 source objects/actions, six products, reference identities
and the catalytic core are preserved. The recursive authored Markdown/YAML
census finds 12 PMIDs, four paper DOIs, two PMC identifiers and three Reactome
records, with no missing source or genuine provider artifact. The legacy JBC
DOI already cited above maps to PMID:6182143 despite the cache lacking a DOI
field: fresh PubMed links to the
exact Elsevier PII S0021-9258(18)33631-7, matching the original article's DOI,
title, authors and bibliographic coordinates. The cache remains unchanged.
The initial census check stopped at this absent metadata field; its explicit
primary-verified mapping resolved the check without inventing a new record.
Full just validate human AK2 passed again with no curation warnings after the
status changed to COMPLETE. The runtime's unrelated dependency-deprecation
notice is not a review-validation warning. History validation and HTML rendering
are recorded in the frozen publication manifest. This session closes the source
gate and leaves PR approval and publication to the coordinator.
The full Results of PMID:24548998 distinguish the major mitochondrial AK2 pool from a nuclear-enriched regulatory pool in HeLa cells. Endogenous AK2-DUSP26 interaction is enriched in that fraction, and forcing AK2 into mitochondria with a COX8 targeting sequence fails to reduce p-FADD. The existing phosphatase-activator proposal now states that experimental context and attaches its exact interaction result. DUSP26 remains the phosphatase; no new nucleus, apoptosis or differentiation annotation is introduced. The sperm mitochondrial-sheath row's copied flagellum wording is corrected, with the mouse donor and human-inference limitation unchanged. All 31 decisions and the catalytic core are preserved.