Falcon deep research was already present and used for the re-review. It supports the core AKT1
function as a PI3K-regulated serine/threonine kinase rather than a proteostasis-specific protein
[file:human/AKT1/AKT1-deep-research-falcon.md "AKT1 is a serine/threonine protein kinase (PKB)"].
The Proteostasis PN projection places AKT1 under chaperone-mediated autophagy regulation and maps it
to GO:1904715 negative regulation of chaperone-mediated autophagy as more specific than the
existing broad GOA annotation to GO:0010507 negative regulation of autophagy
[file:projects/PROTEOSTASIS/reports/pn_projection/pn_projected_gene_go_summary.tsv "AKT1 GO:1904715 negative regulation of chaperone-mediated autophagy more_specific_than_existing_goa"].
The parent PN type is mapped to the directional GO term because it records inhibitory CMA roles
[file:projects/PROTEOSTASIS/mappings/autophagy_lysosome_pathway.yaml "This PN type explicitly records inhibitory roles for CMA"].
The narrower "Modulator of LAMP2A multimerization" subtype is useful triage context but has no direct
GO mapping by itself, so I did not create a LAMP2A-specific GO assertion
[file:projects/PROTEOSTASIS/mappings/autophagy_lysosome_pathway.yaml "curation_status: no_mapping"].
The CMA projection is supported by Arias et al. 2015. The paper identifies a lysosomal
mTORC2/PHLPP1/Akt axis in which Akt1 inhibits CMA; Akt inhibition or Akt1 loss increases CMA reporter
activity and LAMP-2A translocation-complex assembly PMID:26118642.
Mechanistically, the authors connect Akt activity to GFAP phosphorylation and LAMP-2A complex
dynamics, which supports a regulation term rather than direct CMA substrate delivery machinery
PMID:26118642.
Existing GOA already has broad autophagy/proteolysis annotations. The older reporter assay supports
AKT1 as an autophagy inhibitor because AKT1 knockdown increased LC3 reporter release
PMID:18387192.
Macroautophagy evidence remains non-core and context dependent; in mammary epithelial detachment,
PI3K-AKT-MTORC1 activation was not enough to suppress autophagy
PMID:23778976.
Other proteostasis-adjacent AKT1 annotations are substrate or stress contexts, not core functions.
AKT phosphorylation of Rac1 supports FBXL19-mediated ubiquitination and degradation
PMID:23512198.
AKT/CHIP/tau work supports a substrate-specific negative regulation of protein ubiquitination
PMID:18292230.
The heat response annotation is also non-core: Akt suppression increases thermosensitivity, but the
paper attributes acquired thermotolerance primarily to Hsp72/JNK regulation
PMID:10958679.
Curation decision: retain AKT1 core functions as PI3K-regulated kinase activity and PH-domain
phosphoinositide binding. Add GO:1904715 negative regulation of chaperone-mediated autophagy as a
NEW non-core proposed annotation supported by PMID:26118642 and the PN projection. Do not make AKT1
a core proteostasis or direct CMA machinery gene, and do not add a new ontology term.
The starting review contains 445 seeded annotations, one previous NEW proposal,
233 references and two core entries. The current-main baseline was independently
confirmed before editing: review Git blob
2da14e5e9b3cabfd00b96dcf5f924c39f9be8ce6, notes
b5fa9cab0b5eaa035e54ab5ce3e31d61213276c8, and HTML
4ffd4adb26ea99aceb1783a9148e353aa8d63e6e. The parent coordinator found no overlapping
open PR under AKT1, PKBalpha or RACPKalpha. The cached UniProt record is P31749,
with symbols AKT1 and synonyms PKB/RAC; the human nomenclature identifier is
HGNC:391. Machine sources and generated PN notes remain unchanged.
The genuine pre-existing Falcon report (2026-05-02, 486.15 seconds) is retained
as background. It mostly synthesizes reviews and drug-development literature;
its prose is not a replacement for source-specific primary evidence. The first
pass has read the 144 cached PMID abstracts or available source summaries.
Seventy-eight Reactome references are present, of which 24 have local caches and
54 require external source access. All cited PMID paths initially exist, but
PMID:36126419 is only a metadata placeholder and is not a usable abstract cache.
The review is DRAFT while this substantive audit continues.
Scheduling scope correction: the coordinator confirmed that AKT1's strongest
ClinGen association in this campaign is Limited (Cowden syndrome 6), rather than
Definitive. This review was already underway and continues as an explicitly
recorded scheduling exception; the evidence standards are unchanged. The
coordinator owns the corresponding shared-project table and log correction.
The next pass read all 78 referenced Reactome event/pathway records (77 live
event pages, plus the cached CD28 pathway summary and its live Cot-phosphorylation
constituent after the parent page failed). The 136 associated rows now have
source-specific reasons in place of a repeated Falcon sentence. The audit keeps
AKT's kinase activity distinct from its substrates' enzymes and from upstream
kinases acting on AKT. E17K-specific predicted events remain explicitly modeled
inferences; their retention at the general kinase/location level does not claim
that each mutant-specific substrate reaction was experimentally tested.
Five cytosol rows require source-specific plasma-membrane replacements:
R-HSA-1497784, R-HSA-1497796, R-HSA-1497810, R-HSA-202111 and R-HSA-202127.
Their Participants sections place the AKT-containing NOS3/CaM/HSP90 complex at
the plasma membrane; the cytosolic participants are metabolites. By contrast,
R-HSA-202137 genuinely recruits cytosolic AKT into the membrane complex and
supports both locations. R-HSA-377186 currently places the AKT catalyst in the
cytosol and its mTORC1 substrate at the lysosomal membrane, so its old plasma-
membrane row is refined to cytosol. These are source-specific updates, not claims
that AKT lacks either cellular pool. Each live source is linked in its reference
assessment as https://reactome.org/content/detail/<stable_id>.
Reactome R-HSA-199863 explicitly discusses a dispute over the physiological
NR4A1 kinase (AKT versus RSK/MSK). This caveat also affects the predicted E17K
counterpart R-HSA-2399988. It does not negate AKT's independently established
serine/threonine kinase activity or nuclear pool. Several shear-flow event pages
contain inconsistent residue prose; the modeled participants identify AKT
Thr308/Ser473 and NOS3 Ser1177. Original machine-fetched titles remain unchanged.
Initial adjudication replaces 21 legacy generic-protein-binding OVER judgments
with source-specific REMOVE reasons. These remove uninformative molecular-function
assertions without claiming the measured interactions are false. Substrate,
upstream regulator, chaperone-client and scaffold roles are kept distinct. Where an
abstract does not identify the individual curated pair, that limit is explicit.
Other generic-binding rows and all remaining annotation groups are still under
review; existing untouched actions are not yet final audit conclusions.
External primary-source checks completed during this pass:
"PMC2597217" "Akt" "phosphorylation" exposed the same Results and"PMC2658556" "Akt" describe PEBP4The existing CMA NEW proposal also requires reconciliation with the current
nonredundancy rule. The live GO parent page
places GO:1904715 beneath the already annotated GO:0010507. PMID:26118642 supplies
positive evidence for lysosomal Akt control of CMA; the final decision must
preserve that biology while avoiding a redundant new assertion. No CMA action
has been changed in this initial pass.
This completed pass supersedes the initial-pass action statements above and the
older June proposal to retain a CMA NEW row. All 445 seeded assertions, including
their evidence codes, reference identifiers, qualifiers and isoform flags, are
preserved. The one previous NEW descendant proposal is withdrawn, leaving 445
reviewed rows: 204 ACCEPT, 112 KEEP_AS_NON_CORE, 82 REMOVE, 31 UNDECIDED,
15 MODIFY and one MARK_AS_OVER_ANNOTATED. No new assertions were added.
All 233 original reference id/title pairs remain in order; 17 already-cached
donor studies were added, giving 250 assessed references. The two core entries
now cite primary kinetic/activation and lipid-binding evidence instead of a
generated report. The genuine Falcon file and all machine sources are unchanged.
The annotation-reviewer consultation was performed in this lane. A separate
read-only peer consultation independently checked the unusual molecular-function
assertions: Ser/Thr/Tyr kinase, kinase inhibition, self-association and nitric-oxide
synthase regulation. Its source checks and the owner's subsequent reads agree
with the final bounded judgments. The coordinator is independently reviewing the
stable biological draft before publication.
PMID:16540465(https://pubmed.ncbi.nlm.nih.gov/16540465/) directly compares the
three AKT enzymes and identifies ordered ATP/peptide catalysis for AKT1. The
primary insulin/IGF1 experiments in PMID:8978681 test PKBalpha Thr308 and Ser473
mutants and support the central receptor-response mechanism. Phosphoinositide
binding and recruitment are supported by the original
PMID:19203586 abstract, the existing
PIP3/PI(3,4)P2 IDA assertions, and the immutable UniProt domain description.
The original Cell full text was blocked; no full lipid-overlay assay is claimed
as newly recovered from that paper.
The live GO:0004712 definition
requires phosphorylation of serine/threonine and tyrosine. Therefore the two
source-specific replacements by GO:0004674 are chemistry corrections, not a
simple move to a more specific descendant. The RGC-32 source PMID:19162005 tests
Ser45/Ser47 mutants, and the full ZNRF2 source PMID:22797923 maps PKBalpha-dependent
Ser19 by HPLC/Edman sequencing. Neither assay demonstrates target Tyr chemistry;
this does not assert that AKT1 can never show such chemistry in another context.
PMID:23431171 instead directly resolves MOZ Thr369, supporting replacement of
the source's serine-only assertion by Ser/Thr kinase activity.
Superseded scope interpretation: the following two paragraphs read the leaf
definitions without their inherited non-covalent restriction. The 2026-09-27
PR #3242 follow-up below corrects this interpretation; the positive kinase
experiments remain valid.
The literal GO:0030291 definition
requires binding and reduced Ser/Thr kinase activity; it does not impose a
noncatalytic or stoichiometric mechanism. The rat Akt1 accession P47196 carries
IPI evidence to GSK3alpha/beta from PMID:8524413. That primary abstract identifies
PKB as the GSK3-inactivating kinase, and PMID:9373175 directly reports PKBalpha,
GSK3beta Ser9, Ser9A resistance and phosphatase reversal. The full binding assay
was not independently recovered, but the traced donor and conserved human
chemistry support transfer. A pre-existing rat hypothesis report adds a
noncatalytic restriction absent from the GO definition; that inference is not
independent biological contradiction. Rat files were not edited in this scope.
Similarly, GO:0030235
does not exclude catalytic NOS regulation. PMID:10376603 establishes Akt-dependent
eNOS Ser1177 phosphorylation and activation. These regulatory activities are
retained as contextual, non-core functions. Self-association is also retained
with PMID:7891724 and the TCL1 study PMID:10983986; the latter's TCL1 trimers are
not interpreted as an Akt trimer or proof of obligate native Akt dimers.
The following primary routes were read in addition to the first-pass sources.
Indexed PMC queries recovered Results/Methods when direct pages intermittently
returned a browser challenge. They are evidence-access routes, not fabricated
machine caches. Local flags follow actual cache metadata: abstract-only remains
full_text_unavailable: true, whereas a partial retrieved body remains false
with extraction limits stated in the reference assessment.
"Akt-mediated phosphorylation of MICU1" "rapamycin" recovered"PMC1570162" "Akt" "co-localized" recovered Results/Figure 6 using"Spontaneous phosphoinositide" "AktPH" recovered NIH3T3 TIRF experiments"A new cytosolic pathway" PINK1 "Akt" rictor interaction and"PMC3044975" "Akt" "migration" "inhibitor" recovered Figures 2, 3, 5 and 6."PMC3457336" "MC3T3" "RESULTS" recovers relevant preosteoblast PTEN/Akt"PMC12279241" "AKT" "phosphorylation" "TMCO3" recovered Results 2.4-2.7The member index places P31749 in PTHR24356:SF171. The conserved PAINT assertions
were inspected at PTN000682363, PTN000683254 and PTN001219338, using PTN-only
structured source entities. Shared ancestral nodes can occur across AGC family
files; a target's appearance among experimental descendants is expected, not
circular. Donor count is not used as an evidence-strength proxy.
All 29 AKT1 activity records in the cached GO-CAM index were read with their
source-bearing enabled-by, location, process, input and causal edges. They retain
AKT1 as a kinase, with downstream substrates and regulators modeled separately.
Two edge-specific limits are material: cGAS model 62b4ffe300003321 cites
PMID:12172553 on the process edge but PMID:26440888 on substrate/negative-regulation
edges; 65692e7e00001822 combines an RNF115 input with review-derived postsynaptic
potential context. Such combinations prompt source-specific uncertainty rather
than invented assay details. The MICU1 and PERK models preserve their actual MGI
inference routes, independently corroborated by the primary experiments above.
The existing negative-autophagy term covers the earlier proposed CMA child term.
The full PMID:26118642 experiments, including isolated lysosomal Akt/GFAP kinase
assays and LAMP2A complex dynamics, remain positive findings. Withdrawal of the
redundant NEW proposal does not reject this mechanism or add a new process on
necessity evidence alone.
Normal fetch-pmid attempts for all three correction PMIDs failed with DNS
nodename nor servname provided. PMID:36126419 exists only as a 436-byte metadata
placeholder: a normal fetch skipped that existing file, and a supported
temporary-output fetch failed with the same DNS error. No cache was hand-edited
or manufactured. Fifty-four of 78 Reactome sources also lack local machine
caches, despite the verified external event content described above. These
access gates keep status DRAFT; validation success alone does not waive them.
Targeted validation passed with seven nonblocking warnings: six intentional
source-specific action differences (plasma membrane, cytosol, serine kinase,
cGAS regulation, glycogen biosynthesis and glucose import), and the informational
warning that no annotation uses the generated Falcon report as primary support.
This is appropriate: the generated report is retained as background, while
source-specific primary evidence supports the decisions. Final source-preservation,
history validation and rendering are recorded in the handoff manifest.
This entry supersedes the earlier missing-Reactome and metadata-only source
statements. The canonical review, notes, HTML, machine files and prior history
were byte-checked against published PR #3242 head
25067bbc3c0641b650867b75f3d2dac333b84431 before editing.
All 445 source assertions and actions, 250 reference identities, two cores and
existing history are preserved. Only the MPST-source evidence prose/availability
and its reference assessment change; no new annotation is proposed.
All 54 previously absent Reactome records now have normal-fetch caches. Their
stable IDs, nonempty summaries and exact bytes match the verified source2 import
receipt. The summaries corroborate the event-level interpretations already read
on the live Reactome pages; they are not substitutes for the fuller participant
and compartment records previously inspected. In particular, AKT is the substrate
in its activating/deactivating reactions, many E17K substrate events explicitly
remain predictions, and the NR4A1 event retains its disputed physiological-kinase
assignment. The R-HSA-9860759 residue-name inconsistency remains documented in its
reference assessment; the machine title and cached record are preserved.
The normal-fetch candidate for PMID:36126419(https://pubmed.ncbi.nlm.nih.gov/36126419/)
was checked against the original PMC article:
PMID, title, author list, PMCID PMC9486620 and DOI 10.1016/j.redox.2022.102469
match. The old canonical file was a 436-byte metadata placeholder (Git blob
36895e1bd940255b25be03e16fdaa6e81b2744c3). It is replaced by the exact,
unaltered 166524-byte normal-fetch candidate with SHA256
ae05feaf20dcf1275ccc9f87119e4337c7505ecb0d73830dc5020bcec06d8292.
No cache text or metadata was hand-edited.
The recovered Methods 2.13 describe tagged MPST/AKT transfection and
immunoprecipitation in HEK293T cells, plus an assay of recombinant His-MPST and
GST-AKT. Results 3.8/Figure 9 provide the positive interaction and C-terminal
AKT deletion-mapping results. The text generally names AKT, without accession
information resolving every construct or endogenous isoform. The interaction is
retained as evidence while its generic protein-binding annotation remains REMOVE
under the uninformative-term policy. It is not a false-interaction judgment.
A newly visible internal wording inconsistency is retained explicitly. The
abstract says MPST reduces AKT phosphorylation, whereas Results/Figure 8 report
reduced Ser473 phosphorylation after MPST loss and Results 3.8 describe MPST
sustaining AKT phosphorylation. This does not negate the independent Figure 9
interaction experiments. No kinase-inhibitor or other new molecular function is
inferred from that contradictory abstract sentence. The XML extraction repeats
some sections, but includes actual Methods, Results and figure captions; both
review availability flags for this source are now false.
The normal output came from Actions source run 36289953066 at head
fecff1befb769b1753300fa1bc2e3442813e9dd2, transport run 36292331362,
artifact 10923045788. Verified artifact SHA256:
0876942c72b2e537e858e8af7cd3c79d34b97c2169490e3d883c6f00884e2965.
The source2 import receipt records all per-file hashes; the MPST file was held
as a candidate until this identity/content review. GitHub subtree queries confirm
that all 54 Reactome paths are absent from the published baseline and the old
MPST metadata blob is exactly the one replaced.
The three required correction records PMID:33790472, PMID:35267011 and
PMID:36423325 remain pending the coordinated source3 recovery. Their citations
and existing assessments remain intact. This is partial cache closure; the review
remains DRAFT. Final validation, quote/source preservation and rendering results
are recorded in the follow-up manifest.
The normal source3 fetch recovered PMID:33790472, PMID:35267011 and
PMID:36423325. Their exact generated records were imported after ZIP, per-file
hash, source identity and current-main checks. Each cache contains bibliographic
and correction-link information with full_text_available: false; none contains
the notice body. The previously inspected external notices remain the basis for
the correction assessments above. The Nature lane-label correction and the
publisher-provided ZNRF2 notice were independently rechecked; the latter also
states that its separate phosphorylation-site evidence remains sound.
All 164 PMIDs in the finalized review/notes/provider census and all 78 cited
Reactome records are now cached. This closes the machine-cache gaps without
claiming full-text availability for every reference. All 445 annotation decisions,
250 reference entries, two core functions and alternative products remain exactly
as published at 66515f4a322aa8ba282b70c9abdbd6c2cd8022eb. The review remains
DRAFT because its seven explained validation advisories remain; the PR can enter
normal substantive review with no missing source-cache gate. Current-head review
and required tests still govern merging.
Follow-up just validate human AKT1 passed with the same seven explained
source-specific/action and provider-quotation advisories. A separate check finds
94/94 ordinary supporting-text occurrences in their exact local sources using
case-sensitive whitespace-only normalization; no full-text escape fields or
missing quoted sources are present. The notes-inclusive correction-cache gate
remains separate from this successful validator result.
The published baseline was checked at head
2ff0f9d187356afe9d801f751bca34948ddd5005; all seven canonical gene files
matched the remote blobs. Formal review 5329629755 and detailed comment
5854292480 were read in full. The three regulator MF judgments are changed from
KEEP_AS_NON_CORE to MODIFY after checking the actual GO parents, not merely the
leaf labels. All 445 original source assertions, evidence codes, references,
qualifiers and isoforms remain intact; the two cores and all other actions are
unchanged. No NEW annotation is added.
OLS MCP was not exposed in this session. Live official AmiGO pages provided the
definitions and ancestry instead. GO:0098772
restricts molecular-function regulation to “non-covalent binding that does not
result in covalent modification to the target”. Its comment makes the same
restriction explicit. GO:0140678
inherits and repeats that restriction. The live trees place
GO:0030291 under the
inhibitor branch and GO:0030235
under enzyme regulator activity and GO:0098772. Thus the earlier leaf-only
interpretation above is superseded.
The GSK3 evidence remains positive: the cached PMID:9373175 record reports
PKBalpha-dependent GSK3beta Ser9 phosphorylation, loss of inhibition with Ser9A,
and reversal by phosphatase. PMID:21711983 supplies the signaling-review context;
its local abstract does not expose all the reviewed experiments. The traced rat
P47196 IPI source PMID:8524413 and the human corroboration support conserved
kinase-mediated inhibition, not proof of a separately measured non-covalent
inhibitor MF. The complete rat binding assay remains unread, so the donor block
does not claim that an additional non-covalent function is impossible. Similarly,
the actual PMID:10376603 abstract establishes eNOS Ser1177 phosphorylation,
site-mutant failure of activation and a phosphomimetic activity effect. AKT1
performs the covalent phosphorylation step in both cases.
All three rows recommend the established catalytic MF GO:0004674. The live
process definitions GO:0006469
and GO:0051000 describe the
inhibitory kinase and activating NOS outcomes, respectively; they are discussed
as modeling alternatives rather than added as new assertions. Existing kinase
and nitric-oxide-biosynthesis regulation rows retain the measured chemistry and
outcome. Neither the kinase MF nor an ancestor/descendant process is duplicated
with a NEW row.
The MOZ Thr369 replacement stays GO:0004674. Live
AmiGO GO:0004674 explicitly
lists GO:0106311 as an alternate ID and threonine kinase activity among its
synonyms, so the suggested sibling is not a distinct current term in that
release. The original GOA serine-only assertion remains untouched. The brief
description now gives lysosomal functions proportionate contextual emphasis;
TMEM175's supported non-core judgments remain unchanged. Each of the five
Reactome compartment reasons now names the actual AKT1-containing eNOS complex
shown on its live event page. These are source-specific plasma-membrane
assignments, not a denial of AKT1's separately supported cytosolic localization.
Published history record
2026-09-27T061646Z-claude-code-6196e6.yaml is preserved byte-for-byte. Its
filename/session token says claude-code while its actor fields explicitly say
codex, gpt-6 and codex; this mismatch is recorded here rather than silently
rewriting append-only provenance. That record describes closure of 164 PMID
and 78 Reactome cache requirements, with 445 decisions, 250 references and two
cores preserved at that earlier step. The new session is scaffolded with
explicit codex actor/tool and gpt-6 model arguments and identifies the sections
actually changed in this follow-up. No cached source or provider artifact is
edited, and this follow-up adds no literature citation or new source-cache gate.
The full targeted validator passes with the same seven explained advisories
(source-specific action differences and an unused provider quotation). DRAFT
therefore remains appropriate locally. History validation, rendering, exact
source preservation and case-sensitive quote checks are captured in the final
follow-up manifest; current-head CI and independent review remain separate.