ACTA1 (P68133) — review notes

Historical review journal: the 2026-09-26 source and annotation re-audit at the end supersedes earlier judgments, including stress-fiber, interaction-screen and HDA interpretations.
Skeletal muscle α-actin. 377 aa, PANTHER PTHR11937, PE 1: Evidence at protein level.
Reviewed as part of the PAINT + affinage campaign, after ten actin-family siblings
(ACTL7A/7B, ACTL8, ACTL10, ACTR1A/1B, ACTR5, ACTR8, ACTR10, ACTRT3).

Why this gene is the inverse of its siblings

Every actin-family gene reviewed before ACTA1 was a divergent member, where the risk
was inheriting canonical actin biology it does not have. ACTA1 is the canonical
actin, so the phylogenetic transfers that had to be argued down elsewhere are here
correct, and had to be argued up.

The concrete demonstration is GO:0005200 structural constituent of cytoskeleton.
Within PTHR11937 that term is asserted once, at PTN000940351, the
conventional-actin node, by IBD from ten seeds — and negated by IRD at eight
descendant nodes (counted from interpro/panther/PTHR11937/PTHR11937-paint.tsv:
1 IBD + 8 IRD). ACTL7A, ACTL7B and ACTRT3 each needed their GO:0005200 row removed
or generalised. ACTA1 sits on the accepting side of that split and is the recipient
the assertion was made for. All ten of its donors resolve and all ten carry their own
experimental evidence for the term.

The negation sweep ran in two dated batches — five nodes on 2025-08-05
(PTN000233752, PTN000233887, PTN000234048, PTN001732543, PTN008986528) and
three on 2026-04-16 (PTN000233596, PTN000233796, PTN007551901). ACTA1's presence
in the retained set was checked positively, not inferred from the absence of a
negation: a QuickGO query for GO:0005200 in human with evidence ECO:0000318 returns
ACTA1 among the recipients, alongside the other conventional actins (ACTA2, ACTC1, ACTG2,
ACTB, ACTG1, ACTBL2, the POTE paralogues), ACTR10, and the divergent genes not yet
adjudicated. Absence of a negation would have been an absence, not a finding.

So the campaign's reflex — that an IBA over-reaches — is wrong here, and PAINT's node
placement in this family is demonstrably working: it discriminates ACTA1 from ACTL8
correctly. That is worth recording as a positive result, because the family's PAINT
handling has otherwise only been reported when it failed.

The actual defect: three pipelines each substitute a generic actin for ACTA1

ACTA1's over-annotations do not come from disease pleiotropy (there are no
nemaline-myopathy phenotype rows in GOA at all — the 7 BP rows cover only 5 distinct
terms). They come from isoform substitution: three independent pipelines treat
"actin" as one thing, and each injects non-muscle actin biology into a
skeletal-muscle-restricted gene.

This matters because ACTA1's tissue restriction is not a soft claim:
PMID:16288873
— α-skeletal actin is the exclusive sarcomeric isoform in skeletal muscle from
infancy, and α-cardiac actin, not ACTA1, predominates in heart
PMID:16288873.

(a) AgBase ISS — a chicken smooth-muscle actin donates five terms

Rows 19-23 (GO:0010628, GO:0030027, GO:0030175, GO:0044297, GO:0090131) all
have WITH/FROM UniProtKB:P08023, which resolves to ACTA2, actin aortic smooth
muscle, of Gallus gallus.

The donor is not weak — P08023 holds all five terms by its own IDA/IMP — but all five
come from one paper, PMID:10633868, an antisense study of endothelial-mesenchymal
transformation in chick cardiogenesis:
PMID:10633868,
PMID:10633868,
PMID:10633868.

Three things are wrong with the transfer:

  1. Wrong paralog. Chicken ACTA2 is the ortholog of human ACTA2. Chicken has its
    own ACTA1 (P68139 ACTS_CHICK) — the correct ortholog exists in the same species
    and was not used.
  2. Wrong context. GO:0044297 cell body is defined as "The portion of a cell
    bearing surface projections such as axons, dendrites, cilia, or flagella"; ACTA1 is
    a sarcomeric thin-filament protein, not a migratory mesenchymal cell.
  3. ISS is uninformative between α-actins. Measured (see
    ACTA1-bioinformatics/RESULTS.md): ACTA1 is 97.9% identical to ACTA2 and
    98.9% to ACTC1. A sequence-similarity transfer among α-actins is satisfied
    trivially and carries no isoform information — which is exactly why the block
    landed on four genes at once.

Scope, from QuickGO — the block is on all four human α-actins and mouse Acta1:
ACTA2 (legitimate, true ortholog), ACTG2 (arguable, smooth muscle), ACTA1
(sarcomeric — wrong), ACTC1 (sarcomeric — wrong), mouse Acta1 (wrong). The
cytoplasmic actins ACTB and ACTG1 did not receive it. The block is therefore keyed
on α-actin-ness, not on actin-ness, which is the positive control: it discriminates,
and it discriminates on the wrong feature. Retracting it from ACTA1, ACTC1 and mouse
Acta1 fixes three genes in one edit.

(b) Reactome — a generic F-actin (all) polymer

Eight of the 50 GOA rows are GO:0005829 cytosol TAS, one per Reactome reaction. They
are not eight findings; they are one export. Four are Striated Muscle Contraction
reactions (ATP Hydrolysis By Myosin, Calcium Binds Troponin-C, Release Of ADP From
Myosin, Myosin Binds ATP) and one is a dystrophin-glycoprotein-complex reaction —
genuine ACTA1 pathways.

The other three are E-cadherin adherens-junction reactions: R-HSA-9934294
CDH1-associated CTNNA1 binds VCL, R-HSA-9934410 CDH1 forms homotypic trans-dimers,
R-HSA-9934486 CDH1-associated CTNNA1 binds F-actin. ACTA1 is in them because
Reactome's participant is a generic polymer, F-actin (all) [cytosol], whose
reference entities are all six human actins — ACTB, ACTA2, ACTG1, ACTG2, ACTC1 and
ACTA1 (checked via ContentService/data/participants/R-HSA-9934486/referenceEntities).
UniProt's own DR block records the consequence:
[file:human/ACTA1/ACTA1-uniprot.txt "DR Reactome; R-HSA-9764561; Regulation of CDH1 Function."]

So skeletal-muscle α-actin is placed in an epithelial adherens junction by set
membership. The GO term that results, cytosol, is bland enough that the error is
invisible from the GO record alone — which is why it is worth naming.

But the defect is in the pathway membership, not in the GO term, and the actions have
to reflect that. A first draft of this review split the eight rows — KEEP_AS_NON_CORE
for the five muscle reactions, MARK_AS_OVER_ANNOTATED for the three epithelial ones —
and the repo validator was right to flag it: what each of these rows actually asserts in
GO is only that ACTA1 is in the cytosol, and that claim is exactly as true (and as
uninformative) for the CDH1 reactions as for the cross-bridge cycle. Giving an identical
term two different verdicts implied the term was more wrong in some rows than others,
which it is not. All eight are now KEEP_AS_NON_CORE, the routing argument lives in the
three reasons, and the fixable item is filed as a Reactome-side correction in
suggested_questions. Worth recording because the same temptation will arise on any gene
whose over-annotation is upstream of the GO term it produces.

(c) PAN-GO/PAINT — the same substitution running outward from ACTA1

The one row where ACTA1 is the source of the problem. GO:0001725 stress fiber is an
IDA on ACTA1 from PMID:15198992, whose actin-localisation experiments are
transfections: PMID:15198992.
Stress fibers are a property of the undifferentiated C2C12 myoblast host, which
incorporates any actin; the paper's abstract never mentions stress fibers, and the
strong result is in patient muscle
PMID:15198992.
By the campaign's own rule, an ectopic-expression readout is IMP, not IDA.

That single datum then fanned out. PAN-GO promoted it to core: the PAINT file shows
PTN000233075 GO:0001725 C IBD false UniProtKB:P68133 taxon:32523, i.e. one of the six
PAN-GO annotations UniProt advertises for ACTA1 is stress fiber, seeded by ACTA1
itself. And QuickGO shows all four of mouse Acta1's GO:0001725 rows carry
WITH/FROM P68133: ISS (UniProt), IEA (Ensembl), ISO (GO_Central) and IBA (PAN-GO).
One transfection readout became five annotations across two species and four pipelines.

The three sarcomeric siblings on the same PAN-GO node (GO:0005865, GO:0005884,
GO:0030240) are correct and stay; the point is specific to stress fiber.

Peptide-level measurement, and what it does not show

Five rows place ACTA1 outside the cell by high-throughput MS: exosomes from prostatic
secretions (PMID:23533145), parotid gland (PMID:19199708) and trabecular meshwork
(PMID:21362503), plasma microparticles (PMID:22516433) and tears (PMID:23580065). None
sampled skeletal muscle; HPA calls ACTA1 "Tissue enriched (skeletal)".

actin_peptide_specificity.py asks whether shotgun MS can attribute an actin peptide
to ACTA1 at all. Of 63 ACTA1 tryptic peptides in a 7-30 aa window, only 9 (14.3%)
occur in no other human actin, and those 9 collapse to just 3 independent regions
(most are nested missed-cleavage variants — reporting 9 would overstate it). Of the 54
shared peptides, 29 are shared with ACTB and/or ACTG1, which are expressed in every
tissue those five studies sampled.

This does not refute the rows. Three distinguishing regions exist, so ACTA1-specific
attribution is achievable and a study reporting one of those peptides would settle it.
What it shows is that the attribution cannot be assumed — the rows are untested, not
refuted. Deliberately the same distinction ACTL8 drew between its refuted
filament-interface result and its untested ATP-site result.

Interactions: one Y2H screen counted three ways (the ACRV1 pattern again)

Eleven GO:0005515 rows. The fetch-gene stub collapsed seven of them into one; all
seven were restored so each partner gets its own verdict (the ACTR5 lesson —
existing_annotations is counted against the 50-row TSV, not the 44-row stub).

Queried IntAct directly (/intact/ws/interaction/findInteractions/P68133, 142
interactions). All seven PMID:32814053 partners — ZNF20, INCA1, ASCL4, LIAT1, PNMA5,
SYNC, CAMK2A — are logged as two hybrid pooling + validated two hybrid +
two hybrid array, three sub-methods of the same screen, MI-score 0.56, no
orthogonal assay. So UniProt's NbExp=3 is one experiment counted three times, exactly
as found on ACRV1. The panel coheres as a neurodegeneration bait set, i.e. as the
screen's design rather than ACTA1's biology.

One partner deserves a flag rather than dismissal: SYNC (syncoilin) is a genuine
muscle intermediate-filament-associated protein, so an ACTA1–syncoilin interaction is
biologically plausible. It is still a single 0.56 Y2H hit, so it belongs in
suggested_experiments, not in an annotation.

ANXA8 (rows 8, 16) is anti tag coip in BioPlex 2.0 and 3.0 — two datasets, one
method and one pipeline, in HEK293T and HCT116, neither a muscle line.

All ten annotated partners resolve to reviewed Swiss-Prot entries with canonical
lengths; the only non-canonical identifier, Q6ZQX7-4, is a declared LIAT1 isoform,
not a partial ORFeome clone. So the ACRV1 TrEMBL/ORFeome check is negative here.

The asymmetry: GOA has the screens and not the studies

Worth stating separately because it is the one place where ACTA1's record is missing
something rather than carrying too much, and it was only noticed by asking why two
pipeline-fetched publications had ended up uncited.

All eleven GO:0005515 rows come from screens or from incidental observations. UniProt,
meanwhile, records two ACTA1 interactions with experimental ECO:0000269 tags, from
dedicated studies, and neither has any GO row:

Queried by PMID in QuickGO: PMID:10958653 carries 7 annotations in all of GO and
every one is on MYOT or mouse Myot — including MYOT's own GO:0051393 alpha-actinin
binding IDA — so it was curated from the myotilin side and never reciprocated onto actin.
PMID:16501887 carries zero annotations anywhere in GO, despite being titled for
sarcomeric-protein interactions and despite UniProt reading an ACTA1 result out of it.

So the shape of this gene's interaction record is: nine weakly-supported rows from two
high-throughput sources present, two experimentally-supported interactions from
hypothesis-driven studies absent. Adding the latter two would improve the record more than
removing any of the former nine, which is why this is filed as a question rather than as a
set of REMOVEs. (No NEW row is proposed for them, because GO:0005515 is exactly the
uninformative term the project guideline discourages and neither partner has a specific
binding term available.)

Two interactions that are real and stay

What ACTA1 is missing

GO:0051371 muscle alpha-actinin binding is absent from GOA, yet it is one of
ACTA1's best-evidenced molecular functions. UniProt maps it to two discrete segments —
[file:human/ACTA1/ACTA1-uniprot.txt "FT REGION 112..125"] and
[file:human/ACTA1/ACTA1-uniprot.txt "FT REGION 360..372"], both noted
"Interaction with alpha-actinin" — and the affinity has been measured on actin isolated
from human muscle: PMID:16945537,
PMID:16945537.
A mutation that reduces the affinity 10-fold presupposes the wild-type affinity.
Proposed as a NEW annotation.

GO:0006936 muscle contraction (TAS, PMID:10508519) is correct but one level too
general for a gene whose isoform is skeletal-exclusive; GO:0003009 skeletal muscle
contraction is the right term. The cited paper's own framing is skeletal-specific
PMID:10508519, and
the exclusivity is established independently PMID:16288873.

Checks that came back negative (recorded so they are not re-run blind)

Three checks added after the coordinator flagged them mid-review

Notes on the tooling (both cost a cycle, both are guarded now)

Three of my own errors, caught by the repo's own checks

Recorded because each is a general trap, not an ACTA1 quirk.

  1. I wrote a reference title from memory and the pre-write hook blocked it. For
    PMID:16288873 I invented a plausible descriptive title beginning "Comparison of…" and
    naming skeletal and cardiac muscle; the real title begins "Defining alpha-skeletal and
    alpha-cardiac actin expression in human heart and skeletal muscle…" and goes on to
    state the paper's conclusion about the absence of cardiac involvement in ACTA1 nemaline
    myopathy. (The invented string is deliberately not reproduced here: audit_claims.py
    treats it as a retracted phrasing and fails if it reappears on any prose surface.)
    Plausible, wrong, and it is a load-bearing reference — it carries the exclusivity fact behind both the
    GO:0003009 MODIFY and the whole AgBase argument. The same hook also caught
    PMID:23580065, where the cached title spells "naïve" with a diaeresis. Titles must be
    copied from publications/PMID_*.md, never typed.
  2. My dead-accession guard did not work, and its own self-test proved it. I wrote the
    liveness check as "primaryAccession must equal the accession requested", then tested
    it on O15507 — the accession the ACTR10 review found dead. It passed the check:
    an inactive UniProt entry comes back with its own accession and nothing else, no
    protein name, no gene, no organism. The authoritative signal is entryType: "Inactive", and UniProt additionally reports inactiveReason: {MERGED → P56159}. The
    guard now reads that field and names the replacement, and a fifth self-test case
    requires a live accession (P68133) to still pass, so it discriminates rather than
    refusing everything. Note the shape: the accession check was not weak, it was blind to
    the exact class of thing it was written to catch.
  3. A silent zero read as a finding. The first resolver run reported that the four
    C. elegans WITH/FROM donors carry no evidence for GO:0015629. They had never been
    asked: QuickGO rejects every non-UniProtKB gene-product id in this file's WITH/FROM
    fields with HTTP 400, and UniProt does not cross-reference WormBase by WBGene id, so
    both halves of the lookup failed quietly. Fixed by treating 400 as non-retryable,
    re-routing through the resolved UniProt accession, recording the route in the output,
    and verifying every model-organism mapping against the resolved entry's own
    cross-references. The corrected count is 24/24 donors with their own experimental
    evidence    , not 20/24 — and it is the number that makes the GO:0015629 ACCEPT solid.

What PR review caught (two factual, three presentational)

The reviewer approved with no blocking findings, and two of its five suggestions were real
errors that I verified before conceding:

  1. The peptide analysis digested the ORF, not the protein. ACTA1 has INIT_MET 1 "Removed" and CHAIN 3..377, the acetylated Cys-2 of the intermediate form being
    cleaved by ACTMAP, so the ORF's N-terminal tryptic peptide does not exist in vivo — and
    my suggested_experiments entry named it, and gave the region as starting at residue 1
    rather than 3. The observable peptide is DEDETTALVCDNGSGLVK, region 3-30.
    (The two retracted strings are deliberately not reproduced on any prose surface;
    audit_claims.py fails if either reappears.) Recomputed on the mature
    chain, with comparators contributing both their ORF and mature digests: all counts are
    unchanged     (63 / 9 / 14.3% / 54 / 29 / 3 regions), and the script now asserts that the
    two forms agree. So the numbers were never wrong; the experiment was unbuildable. The
    general shape is worth keeping — a sequence analysis is only as biological as the
    sequence it starts from    , and a UniProt CHAIN feature is where that is decided.
  2. A by-similarity residue assignment stated as fact. The description gave the two
    MARTX cross-link residue positions as though they had been measured on ACTA1;
    UniProt tags both CROSSLNK features ECO:0000250|UniProtKB:P60709, i.e. transferred
    from beta-actin. The cross-linking itself is real, the residue numbering on the skeletal
    isoform is not measured, and the description now says so.

The other three were presentational and all three were taken: GO:0015629 now appears in
core_functions.locations and its row explains why the child is accepted while the bare
GO:0005856 parent is demoted; the GO:0051371 row explains why the evidence code is IDA
rather than IPI (the paper names only "the Z-line protein alpha-actinin", so there is no
partner accession to put in a with/from field, and inventing ACTN2 or ACTN3 would be a
guess); and the GO:0043531 reason now leads with the nucleotide cycle — ATP-bound G-actin
is the assembly-competent species, the ADP protomer is the aged form that gives the
filament polarity and that cofilin recognises — with the cross-review precedent demoted to
corroboration.

A third error of my own: I relayed a sibling's claim as a fact

Found while auditing my own numbers after the review round, and the most instructive of the
three because nothing flagged it — it was internally consistent, verbatim-sourced and wrong.

The GO:0043531 row placed ACTA1 among a supposed pair of family members holding both
nucleotide terms. That came from the merged ACTR10 review, whose line reads
GO:0005524 ATP binding: ACTA1, ACT1 — a list about ATP binding alone, which I turned
into a count about two terms. Measured across all 533 reviewed (Swiss-Prot) PTHR11937
members: 31 reviewed members carry GO:0005524 and exactly 1 carries GO:0043531 —
ACTA1. It is the sole ADP binding holder among reviewed entries, not one of two, and the
corrected fact is stronger than the wrong one.

The scope qualifier is not a hedge, and PR review was right to insist on it. That member
list comes from InterPro's reviewed-only endpoint, so 533 is ~0.6% of the 88,887 proteins
PTHR11937's metadata reports. An earlier draft called it the whole family, which the
measurement does not support. The family-wide version is an argument and is now labelled as
one: ACTA1's GO:0043531 is manual TAS, unreviewed entries get only IEA, and no IEA pipeline
maps the actin fold to ADP binding — so an unreviewed holder is unlikely, but unlikely is not
measured.

The brief's rule is relay a sibling's claim as a claim, not as a fact, and I broke it in
the specific way that is hardest to catch: a coordinator's or sibling's summary carries more
apparent authority than the review it came from, and this one arrived pre-formatted as a
quotable line.

Getting the right answer took two attempts, and the first failure is the same shape as the
WormBase one earlier in this review: querying by GO term alone returns page 1 of ~205,000
(GO:0043531) and ~9.6 million (GO:0005524) annotations, and intersecting that page with
the family yields an empty set for both terms — which reads as "no member carries this",
the exact opposite of the truth. nucleotide_terms_in_family.py keys the query on the
family's accessions instead, batches them, and asserts no batch was itself truncated.

A fourth error, and the reviewer found it in a guard I had just written

The second review approved but left three notes on the new script code. All three were
real, and two were defects in guards — the class this review had already tripped over twice.

  1. orf_and_mature_counts_agree compared cardinalities, not sets. Fixing it
    immediately exposed a live disagreement the count had masked: both forms yield exactly
    9 distinguishing peptides, but the sets differ in 4 members. So the cheap check
    was certifying an agreement that did not hold.

The right invariant turned out to be narrower than "sets agree", because the sets are
not supposed to be identical — modelling the N-terminal processing is the whole point, so
the N-terminal peptides must differ. What is asserted now is: the counts agree and every
peptide the two forms disagree about lies at the N-terminus. A divergence anywhere else
would mean the offset had corrupted the digest.

  1. And then that guard did not fire when I broke it. Setting MATURE_START = 50 should
    have been rejected; instead the script produced a confident, wrong region (50-64 GQKDSYVGDEAQSKR) with no complaint. The cause: I had written the tolerance as
    MATURE_START + 1, so it scaled with the parameter under test — a bigger offset bought
    itself a wider window. A guard whose tolerance is set by the thing it guards is worse than
    no guard, because it still reports success.     The tolerance is now a fixed constant grounded
    in the biology (N-terminal processing removes at most a couple of residues), and
    MATURE_START = 50 is rejected outright.

Break-testing further showed an honest limit: MATURE_START = 4 passes every check and
yields region 4-30. The guard distinguishes a corrupting offset from a plausible one,
not a correct one from an off-by-one. So the number is no longer typed in at all —
mature_chain_start() parses it from the CHAIN feature in ACTA1-uniprot.txt whose note
matches the entry's RecName, and fails if that is not exactly one feature. It returns 3,
agreeing with the value I had hand-typed; the point is that it is now derived rather than
transcribed. Any constant read out of prose and typed in is a latent bug.

  1. MATURE_START = 3 was documented for ACTA1 but reused for all six comparators. ACTB
    is annotated CHAIN 1..375 and CHAIN 2..375 "N-terminally processed", so its
    observable forms begin at residue 1 or 2 and slicing it at 3 both mis-stated its processing
    and missed a form. Comparators now contribute their digests at offsets 0, 1 and 2, which
    only enlarges the comparator pool and so only shrinks the distinguishing set — conservative
    by construction, and independent of any per-gene annotation being complete.

  2. The reviewer also found a hole in the claim lint itself: a retracted phrasing had
    survived in a post-mortem because an embedded quotation mark split the matched substring.
    The guard was evadable by punctuation while still reporting OK. flatten() now strips
    quotation marks and normalises dashes, and it caught the surviving instance on the next
    run.

Every headline number survived all of this unchanged: 63 peptides, 9 distinguishing (14.3%),
54 shared, 29 shared with ACTB/ACTG1, 3 regions at 3-30, 287-317 and 338-361.

Action summary (50 GOA rows + 1 NEW)

action n what
ACCEPT 15 thin filament / sarcomere / actin filament CCs, GO:0005200 (×2), ATP + ADP binding, ATP hydrolysis, myosin binding, thin-filament assembly (×2), actin cytoskeleton (×2)
MARK_AS_OVER_ANNOTATED 16 9 bare GO:0005515 (7 Y2H + 2 BioPlex ANXA8), 5 extracellular HDA, both GO:0001725 stress fiber rows
KEEP_AS_NON_CORE 13 GO:0005856, 2× GO:0048741, all 8 GO:0005829 Reactome cytosol, the DNASE1 and HBHA interactions
REMOVE 5 the chick smooth-muscle-actin ISS block
MODIFY 1 GO:0006936 → GO:0003009 skeletal muscle contraction
NEW 1 GO:0051371 muscle alpha-actinin binding

Counts are read off the finished YAML, not tallied by hand; 50 non-NEW rows against 50 GOA
rows, asserted by build_source_entities.py.

2026-09-26 source and annotation re-audit

This section supersedes the earlier annotation judgments and biological interpretations in this journal. The earlier record is retained as provenance, including decisions now corrected. The saved bioinformatics outputs remain historical computational results, not evidence that a particular experiment observed a shared peptide or that an interaction was false.

Baseline, scope and research access

Primary recovery changed the stress-fiber judgment

PMID:15198992, Ilkovski et al., was retrieved as the original published paper from the University of Geneva archive, DOI 10.1093/hmg/ddh185. The successful normal curl download is /tmp/ACTA1-PMID15198992.pdf; pdftotext -layout produced /tmp/ACTA1-PMID15198992.txt. The web reader initially exposed the article but later timed out; the complete local PDF recovery resolved that limitation.

Results, journal page 1733, and Figure 5A show wild-type ACTA1-EGFP in C2C12 myoblast stress fibers: “stress-fibre or diffuse cytoplasmic staining”. Figure 5B(viii) and page 1734 show striated incorporation after differentiation. Thus both existing stress-fiber annotations are retained as KEEP_AS_NON_CORE, reflecting the observed culture context relative to mature sarcomeric function. Ectopic expression does not itself make a localization experiment IMP rather than IDA. The earlier abstract-based rejection and claim that stress-fiber incorporation was only mutant behavior were unsupported. Patient-muscle fractionation also supports the cytoskeleton and thin-filament assertions, while its insoluble fraction need not consist exclusively of normal sarcomeric filaments.

Interaction evidence and the generic-binding policy

PMID:32814053 was retrieved as the original full published article from the MDC repository, DOI 10.1016/j.celrep.2020.108050. Genuine artifacts: /tmp/ACTA1-PMID32814053.pdf and /tmp/ACTA1-PMID32814053.txt. Results pages 2–4 report four independent screens, pairwise retesting and DULIP validation of a subset. The prior claim that three interaction-method labels established one experiment counted three times is withdrawn. This recovery does not establish whether any particular ACTA1 pair was included in the DULIP subset.

The seven source pairs (ZNF20, INCA1, ASCL4, LIAT1 isoform 4, PNMA5, SYNC and CAMK2A) remain intact. Their GO:0005515 rows now use REMOVE because the term supplies no informative molecular function; the physical-interaction reports are not declared false. The same policy applies to the two ANXA8 BioPlex rows (PMID:28514442, PMID:33961781) and the genuine DNase I affinity-purification interaction (PMID:12849983). Neither a non-muscle cell line nor a hypothetical actin carry-over establishes contamination. Binding to DNase I does not by itself demonstrate inhibition, so no inhibitor activity is added.

PMID:18835984 differs because the exact gene-product assignment remains unresolved. The original PMC article, Methods, preparation of actin-modified surfaces, specifies bovine-muscle actin. This positively supports the reported HBHA interaction with muscle actin but does not establish a direct human ACTA1 assay. Because the full source is available, the generic-binding specificity policy can be applied: the human IPI row is REMOVE solely for being uninformative, while its direct human attribution remains explicitly unresolved. This is not a declaration that the underlying binding is false, and no specific ACTA1 activity is invented. The zero-evidence assumption from the old abstract-only reading is replaced by the actual reagent provenance.

PMID:10958653 is left UNVERIFIED for its ACTA1-specific interaction interpretation: the cached abstract describes myotilin and alpha-actinin. UniProt's ACTA1 interaction statement alone does not demonstrate that curators deliberately omitted a reciprocal annotation. PMID:16501887 does explicitly identify ACTA1 among USP25m-associated proteins; no USP25 catalytic or proteolysis activity is transferred to its binding partner.

Extracellular HDA assignments remain unresolved

The five original HDA records concern prostatic-secretions/urine exosomes (PMID:23533145), plasma microvesicles (PMID:22516433), tears (PMID:23580065), parotid exosomes (PMID:19199708), and trabecular-meshwork exosomes (PMID:21362503). Each is now UNDECIDED.

The main text was read for the three full-cache exosome studies. It points to supplementary protein lists; the ACTA1-specific protein/peptide entries were not recovered. The two remaining caches are abstract-only. Attempts to access PMC supplement material encountered browser gating, unavailable downloads or DNS failure. The parotid paper identifies supplementary tables 1 and 2, but the file contents were not inspected; an indexed file name is not peptide evidence. The trabecular-meshwork main list includes other actins, which neither proves nor disproves ACTA1 in the full list.

The preserved peptide analysis establishes that ACTA1 has both shared and distinguishing tryptic peptides. It does not identify which peptides were measured in any of these studies. Tissue enrichment, no classical secretion signal, and possible shared peptides are reasons to inspect source data, not grounds to overrule HDA curators. Likewise, absence of a distinguishing peptide in a future reanalysis would leave ambiguity rather than alone prove that ACTA1 was absent.

Phylogenetic provenance and source compartment resolution

The cached PAINT table interpro/panther/PTHR11937/PTHR11937-paint.tsv was checked for the IBDs at PTN002631484, PTN000940351 and PTN000233075. PTNs are ancestral assertions, not irrelevant entries merely because they are not extant proteins. Their structured status now supports transfer, and IBA source_entities contain only the verified PTN proximate assertions; the extant evidence list remains in the unchanged underlying source records. ACTA1 appearing among experimental descendant sources is expected and is not circular. The retained structural-function IBD and its descendant loss annotations do not license an invented label for the entire clade. No acceptance decision is based on the number of extant donors.

No ACTA1/P68133 entry was found in the local GO-CAM index. The existing source-resolution JSON identifies pig skeletal-actin P68137, mouse Acta1 P68134 and chicken smooth-muscle ACTA2 P08023. The additional Ensembl protein token on the Compara row was preserved as UNRESOLVED: its exact cross-reference was not independently recovered, and it is not dismissed as “not a gene product”.

The cytoskeleton IEA now remains ACCEPT at the location-mapping source's resolution. The four cytosol TAS records from actual actin-myosin contractile events (R-HSA-390593, R-HSA-390595, R-HSA-390597, R-HSA-390598) also remain ACCEPT at the pathway's compartment resolution. The event names identify myosin or troponin chemistry, which is not attributed to actin. The four adhesion-context cytosol records (R-HSA-9914537, R-HSA-9934294, R-HSA-9934410, R-HSA-9934486) remain KEEP_AS_NON_CORE: the broad location is compatible, but generic actin-set participation does not establish an ACTA1-specific adhesion mechanism. Thus the differing cytosol actions reflect source context, not an allegation that the compartment is wrong or a rule that every duplicate must be demoted.

The five ISS rows from chicken ACTA2 (P08023) are UNDECIDED. PMID:10633868 expressly studies smooth-muscle alpha-actin in chick endocardial mesenchyme. Its source experiment is valid; conservation of these developmental/localization assertions in ACTA1 is unresolved. Independent annotation-reviewer consultation correctly identified that paralogy, tissue context and missing target-specific evidence do not alone demonstrate transfer failure. The recovered non-sarcomeric ACTA1 localization further argues against an absolute compartment-exclusion claim. An ISS may legitimately cross paralogs; positive evidence of conservation or target-specific divergence is needed to settle these rows. Mouse Acta1 developmental rows remain non-core; PMID:2731651 is expression analysis of an actin-gene-pair perturbation, not a targeted Acta1 knockout.

Ontology and core chemistry

A live QuickGO JSON request succeeded on 2026-09-26; response saved as /tmp/ACTA1-GO-terms.json:

https://www.ebi.ac.uk/QuickGO/services/ontology/go/terms/GO:0051371,GO:0044297,GO:0001725,GO:0003009,GO:0005200

PMID:24743229 was read in full. Its canonical biochemical control is endogenous pig skeletal-muscle alpha-actin (Methods), and Figure 6D compares phosphate release. The cached Results state: “As expected, α-actin showed an even lower release of phosphate in the Ca2+-bound compared to the Mg2+-bound form.” The paper cautions that phosphate release can lag ATP cleavage. This result-bearing support replaces the prior title-only core quote. Hydrolysis influences filament dynamics; oriented subunit polymerization establishes structural polarity. The actin ATPase is distinct from the myosin ATPase powering contraction. No claim is made that human ACTA1 kinetics have never been measured anywhere.

The description now includes dominant and recessive disease mechanisms, uses predominant skeletal-muscle expression rather than absolute tissue exclusion, and omits transferred PTM details from its core biological synopsis. PMID:23673617 centers on beta-actin/ALKBH4, and PMID:30626964 concerns SETD3-dependent actin modification; neither makes ACTA1 the modifying enzyme.

Current action totals and verification

Action Entries
ACCEPT 20
REMOVE 11
KEEP_AS_NON_CORE 8
UNDECIDED 10
MODIFY 1
NEW (retained prior proposal) 1

Totals are 51 entries, including 50 unchanged source assertions. The ten UNDECIDED entries are the five extracellular HDA records and five ACTA2-source ISS transfers. All eleven generic-binding rows are removed for lack of functional information; the HBHA human-protein assignment remains unresolved in the source interpretation. All primary quotes were checked against the immutable cache; external full-text evidence is identified by source URL and section. The Figure 5A excerpt uses supporting_text_fulltext, separately checked against the genuine recovered PDF text and not represented as a cached-publication quote. Source comparison and trailing-space removal both assert parsed equality outside authored review fields. The old bioinformatics claim lint encodes earlier action counts and biological conclusions; it is preserved as historical tooling, not used to force the current review back to obsolete judgments.

Verification completed: just validate human ACTA1 passed with two nonblocking warnings (source-specific cytosol action split and unused provider support). just validate-history passed for the newly scaffolded session, and just render human ACTA1 succeeded. Independent annotation-reviewer consultation read all entries, four cores and 38 references; its two substantive provenance/transfer concerns were resolved and it reported no remaining biological blocker. No Git or remote mutation was performed.

2026-09-26 follow-up to PR #3163: transfer scope and evidence attachments

This entry supersedes the preceding cytosol action split and clarifies the five unresolved chicken-source transfers. All 50 seeded assertions and the pre-existing alpha-actinin-binding proposal retain their source fields. The existing computational scripts, JSON outputs, RESULTS.md, publication caches and published history are unchanged.

Chicken ACTA2 transfer: observations retained, conservation unresolved

The saved withfrom_resolution.json identifies P08023 as chicken ACTA2 and traces four IDA annotations (gene-expression regulation, lamellipodium, filopodium and cell body) and the migration IMP to PMID:10633868. The primary PubMed record was read again on 2026-09-26; the linked Wiley full paper remained inaccessible. The cached record is abstract-only. Its abstract directly describes punctate SMA in lamellipodia/filopodia and antisense effects on mesenchymal formation and migratory appendages, but does not resolve the full gene-expression or cell-body assays. Each ISS reason now addresses its own donor observation and unresolved transfer mechanism; stress-fiber incorporation is not evidence for filopodia, lamellipodia or mesenchyme migration.

All three arguments in the historical analysis were considered. First, the available chicken ACTA1 ortholog P68139 establishes that P08023 is a paralog source. An ISS asserts sequence-based inference and can cross paralogs; the existence of an ortholog does not by itself demonstrate that the paralog's function was lost in ACTA1. Second, the saved distribution across muscle actins, excluding ACTB/ACTG1, and the high ACTA1/ACTA2 sequence identity are retained observations. They describe the transfer's scope and similarity, but contain no assay of conserved regulation, targeting or migration. They justify examining the transfer closely rather than establishing its biological correctness or failure. These computational observations were not refuted or regenerated; the inference drawn from them is what remains unresolved.

Third, live AmiGO GO:0044297, read on 2026-09-26, defines the nucleus-containing portion of a cell with projections, excluding the projections, and cautions about cells without projections. Mature skeletal-fiber morphology therefore cannot positively support this location. It also cannot exclude ACTA1 in every other cell state. The existing annotation has no mature-fiber restriction, and the full donor assay and relevant conservation evidence are unavailable. The cell-body row therefore remains UNDECIDED without asserting that stress fibers demonstrate a cell body or that the term applies to every cell.

UNDECIDED describes unresolved biology as well as enforcing the mandatory rule for inaccessible relevant publications. A cached abstract is not full access to every assay underlying an annotation. The reasons now state the actual evidence examined and what it fails to resolve; they do not treat all ISS as protected from challenge.

Extracellular attribution, cytosol and nucleotide evidence

Each of the five HDA rows now cites the existing RESULTS.md numerical peptide analysis: 54 of 63 theoretical mature-chain peptides are shared with another human actin, with 29 shared with ACTB/ACTG1 and three independent distinguishing regions. These are predicted digestion products, not observed peptides from the five studies. The actual supplementary assignments remain unavailable, so the rows remain UNDECIDED without inferring contamination or false isoform attribution.

All eight Reactome cytosol annotations are now ACCEPT. Cytosol is the broad compartment containing ACTA1's core intracellular filament machinery and is explicitly included in the structural core locations. The four adhesion-context rows now attach their own event quotations and the independently recovered wild-type ACTA1 cytoplasmic localization. Acceptance concerns this location; neither generic actin-set membership nor the event title establishes an ACTA1-specific adhesion mechanism. The contractile event chemistry remains assigned to myosin or troponin. Obsolete row-index/action comments were removed.

Both ATP and ADP binding now cite the actual pig skeletal-alpha-actin control in cached full PMID:24743229, including Figure 7's ATP/ADP comparison. The experimental species is explicit and human affinity measurements are not implied. This replaces an EC-only ATP-binding excerpt and restores the absent ADP evidence attachment. Source-species provenance remains in the ATPase knowledge gap rather than its biological description. Thin-filament assembly reasoning now states positive structural participation without rebutting a prior draft.

The two recovered PMID:15198992 excerpts are both exact and occur at distinct places: Results, journal page 1733, says “stress-fibre or diffuse cytoplasmic staining”; Figure 5A's caption, page 1735, says “diffuse cytoplasmic staining or stress-fibre staining for WT-actinEGFP.” Both were checked against /tmp/ACTA1-PMID15198992.txt, extracted from the genuine institutional PDF. The YAML uses the caption excerpt; their different order is not a quotation mismatch.

Checks and historical tooling

The existing audit_claims.py was actually run. It fails because it mandates old literal phrasings and repetition across prose surfaces, including numeric sequence-identity wording, a particular distinguishing-peptide sentence, a digit-form region count, a spelled-out shared percentage, the old interaction score, and reviewed-family numerator/denominator mentions in the review YAML. It is not reported as passing. The script and its computational inputs remain preserved; no analyses were rerun and no curation narrative was padded to satisfy obsolete literal requirements. The action tally below does reflect the current YAML. The final log is /tmp/acta1-followup-legacy-lint.log; standard repository validation and source-preservation checks are reported separately.

Action Count
ACCEPT 24
REMOVE 11
KEEP_AS_NON_CORE 4
UNDECIDED 10
MODIFY 1
NEW 1

The only action changes are the four adhesion-context cytosol rows from KEEP_AS_NON_CORE to ACCEPT. The specific pre-existing NEW muscle alpha-actinin binding annotation remains unchanged. No new process, molecular-function or localization annotation is proposed.

Final follow-up checks: just validate human ACTA1 passes with one existing unused-research-report advisory; the previous cytosol action-split advisory is resolved. just validate-history passes for the new codex session and just render human ACTA1 succeeds. All 51 source objects (including the existing NEW proposal), alternative products, reference identifiers/titles, machine-source files, computational artifacts and prior history are preserved. The final legacy lint reports seven literal-prose requirements, with no remaining action-tally mismatch; this separate historical-tool limitation is disclosed rather than treated as a standard validation failure or silently repaired by rewriting historical artifacts.