ACAD8 (Isobutyryl-CoA dehydrogenase, mitochondrial) — review notes

UniProt: Q9UKU7 (ACAD8_HUMAN). HGNC:87. Gene synonyms ARC42, IBD. EC 1.3.8.5.

Core biology (from UniProt Q9UKU7 and cited primary literature)

ACAD8 is the mitochondrial isobutyryl-CoA dehydrogenase (IBD/IBDH), an FAD-dependent
member of the acyl-CoA dehydrogenase (ACAD) family. It catalyzes the third step of
valine catabolism: the FAD-dependent alpha,beta-dehydrogenation of isobutyryl-CoA
(2-methylpropanoyl-CoA) to methacrylyl-CoA, with electrons
passed to the electron-transfer flavoprotein (ETF).

Identity/discovery

Disease — Isobutyryl-CoA dehydrogenase deficiency (IBDD; MIM:611283)

Historical ARC42 association (not in the current annotation seed)

The fetched UniProt record links ACAD8/ARC42 identification in the ARC/DRIP preparation to
PMID:10235267(https://pubmed.ncbi.nlm.nih.gov/10235267/) and reports possible complex membership.
The publisher abstract describes the multisubunit complex's transcriptional activity, but the full
article was not accessible in this audit. The current 23-row GOA seed contains no transcription
annotation. The earlier note claiming a current DNA-templated-transcription IEA was therefore
incorrect for this source snapshot. No NEW transcription annotation is proposed, and no categorical
claim is made that ACAD8 lacks a context-dependent role based only on an inaccessible full paper.

Interactions / proteomics annotations

GO term notes

2026-09-26 complete annotation re-review

The existing review matched GitHub main blob e357e75aef6b7754df2a6134d1619d14d2639403, and no
open ACAD8 pull request was found. Its status was INITIALIZED despite 23 already adjudicated rows.
All 23 original annotation field sets were preserved. The updated decisions are 11 ACCEPT,
10 MODIFY, 1 KEEP_AS_NON_CORE and 1 REMOVE, with nine source-entity propagation audits and no NEW
annotations. The status is COMPLETE for this documented manual audit.

The PAINT source is ancestral node PANTHER:PTN000097535; the target's own experimental evidence
among descendants is not circular. Rhea reactions 44180 (isobutyryl-CoA), 48256
(S-2-methylbutyryl-CoA) and 31287 (propionyl-CoA), EC1.3.8.5, UniPathway UPA00362, InterPro domains
and UniProt subcellular-location mappings were traced against the untouched GOA and UniProt
records. ARBA and precise InterPro mapping internals were not inspected and are explicitly UNRESOLVED rather than
represented as independently validated.

Live ontology pages verified the relevant substrate-class scopes:
GO:0003853 and
GO:0016937. The latter permits branched
substrates and is the former's direct parent. The retained propionyl side-reaction rationale is
therefore about its particular Rhea source, not a straight-chain restriction in GO.

Broad mitochondrial annotations now refine to the matrix rather than being labeled non-core just
because they are broad. All five matrix annotations are accepted as a core location. The four
proteostasis Reactome event narratives discuss matrix substrates without naming ACAD8 individually;
the exact substrate membership was not independently reconstructed. The location is supported
independently by its explicit catalytic reaction R-HSA-70859 and human import assays. Candidate
substrate status does not give ACAD8 the protease's function. No degradation or proteostasis process
was added. MitoCoP's cached main text does not expose the ACAD8 supplementary row; this absence
was not treated as evidence against the curator's experimental localization assignment.

The PMID:12359132 abstract's mitochondrial-targeting sentence specifically concerns Arg302Gln,
which retains targeting despite lost activity. The wild-type import evidence is explicit in
PMID:11013134. These source contexts are now distinguished. The latter paper directly assays
ACAD8 despite a title emphasizing the ACADSB defect. Four core experimental/sequence papers are
abstract-only in cache; no inaccessible full-text experiment was claimed as newly inspected.
The BioPlex cache is labeled full_text_available but contains selected narrative sections; the
specific partner comes from GOA/UniProt/IntAct, not the study-wide quoted abstract.

FAD binding is retained as an annotation and integrated into the dehydrogenase core function,
rather than presented as a second independent biological role. The structure paper's public
publisher abstract supports
the homotetramer, FAD/product geometry and substrate-pocket interpretation. Its full text was not
retrieved; specific structural residue claims are not added. The main biological summary avoids
unnecessary clinical prognosis and uses methacrylyl-CoA without importing a stereochemical prefix
from a generalized chemical-class label.

Research and access provenance

The required default Falcon workflow and GOA publication caching were launched concurrently.
All six seeded PMID caches were already present and confirmed by the normal fetch workflow.
The first genuine Falcon/fallback command failed before contacting either provider because uvx
could not create a temporary directory under the read-only user tool-install path (OS error1).
A coordinator-authorized retry used process-local writable UV_TOOL_DIR, UV_TOOL_BIN_DIR and
UV_CACHE_DIR paths under /tmp; this corrected that local path issue. Both provider attempts
then failed while resolving pypi.org for deep-research-client, after uv's internal dependency
retries (provider process exit2, wrapper exit1). No provider report was produced, and none was
created manually. No global configuration or permissions were changed.

Normal attempts to cache additional PMID:14752098 (structure) and PMID:10235267 (ARC) also failed
with urlopen error [Errno 8] nodename nor servname provided, or not known; 0/2 new publication
caches were created. Public primary abstracts were available through web search, but direct
publisher/full-text routes were inaccessible or provided only abstracts. These papers remain
background/structural context through existing UniProt provenance and the notes; the reviewed
source annotations retain their six available primary caches and five cached Reactome sources.
No source file or cached publication was hand-edited. No Git or shared project changes were made.

Logs are /tmp/ACAD8-falcon.log, /tmp/ACAD8-falcon-writable-runtime.log,
/tmp/ACAD8-publications.log and /tmp/ACAD8-additional-publications.log. Validation, rendering,
independent review and the exact publication manifest are recorded with the new history session.

The coordinator independently read all 23 annotation decisions and the core synthesis, agreeing with the biological judgments. Final citation cleanup grounds the FAD row in the available cached UniProt cofactor/structure record; the unsuccessful structural-paper fetch does not become a claim of additional primary full-text coverage. The manual full-text-unavailable flag on BioPlex reflects the incomplete narrative extraction, without altering its cached metadata.

2026-09-26 PR #3151 evidence-scope follow-up

The external review identified over-refinement of five mitochondrial-location
source rows. They now ACCEPT mitochondrion at the original evidence resolution;
the separately supported mitochondrial-matrix core and five Reactome matrix rows
remain unchanged. Import assays, MitoCoP and the PAINT mitochondrial node are not
represented as direct matrix-localization experiments.

Rechecked the live GO:0016937 hierarchy:
GO:0003853 is its direct child. The broad GO activity therefore encompasses ACAD8's
core isobutyryl reaction and is now ACCEPT, while RHEA:31287 retains its narrower
propanoyl-to-acryloyl side-reaction provenance. Replacing that source reaction with
2-methyl chemistry would be incorrect. Independent biological consultation agreed
that the action should grade the asserted GO term while the reason preserves the
specific source reaction. Older sibling-review wording that describes these GO
terms as siblings is not current ontology evidence.

The lipid-to-valine replacement is explicitly a cross-branch pathway correction.
Removed the full_text_unavailable flag from the BioPlex reference: its missing
pair-specific supplementary evidence is documented separately from its available
article narrative. The transcription question now asks for a biological mechanism.
Uninspected InterPro rule details remain UNRESOLVED: source identifiers and their
protein-domain labels are recovered, but a sibling review's acceptance is not an
independent reconstruction of each mapping rule. ARBA predicates remain opaque.

All 23 original source-field sets are unchanged. Final actions: 17 ACCEPT,
5 MODIFY and 1 REMOVE; no NEW annotations. No cached source or provider file was edited.