Human short/branched-chain specific acyl-CoA dehydrogenase, mitochondrial (SBCAD;
2-methylbutyryl-CoA dehydrogenase; ACADSB). HGNC:91, gene ID 36, chromosome 10.
UniProt reviewed entry P45954; 432 aa precursor with an N-terminal mitochondrial
transit peptide (residues 1..33) cleaved to give the mature matrix enzyme.
This entry supersedes the earlier annotation conclusions above where they differ. The prior journal is retained. The scope was all 28 existing assertions (no NOT annotations and no proposed NEW rows), the three previous core-function entries, all five cached PMIDs, and all propagated-source identifiers. Seeded assertion fields and both alternative-product records were preserved. The resulting review contains 24 ACCEPT and four KEEP_AS_NON_CORE decisions, with one integrated enzymatic core.
ACADSB is the canonical human symbol (HGNC:91; UniProt P45954; NCBI Gene 36). HGNC-attributed ClinGen gene facts list SBCAD and ACAD7; Ensembl also exposes 2-MEBCAD. This is distinct from ACADS, ACAD8 and ACADM. The remote main revision checked was 21121fc735d20bcb2dbf8328aa82d5da30185e5e; all five local ACADSB files matched their remote main blob identifiers. A GitHub open-PR search for ACADSB returned zero results at the baseline check. The initial YAML blob was 691bbbbc55ba2035e8eff56408601c1179b86879 (status INITIALIZED). Baseline files and hashes were retained separately for source-tuple and publication integrity checks.
The default genuine Falcon attempt ran with a 1200-second requested timeout and automatic perplexity-lite fallback, concurrently with publication caching. Both provider paths failed while bootstrapping deep-research-client[cyberian]==0.2.7rc1: PyPI DNS resolution failed after three retries. The Falcon backend and the fallback backend were never reached. This was neither a completed literature report nor a service timeout/quota result. No provider-named report was written or invented. fetch-gene-pmids completed with all five existing papers already cached and unchanged. A separate supported forced fetch of 11013134 and 12855692 to a temporary output directory failed DNS for both. The review therefore used cached material and directly inspected primary web sources; machine-owned cache files were not edited.
The principal physiological reaction is (S)-2-methylbutyryl-CoA to tiglyl-CoA in isoleucine degradation. The human recombinant-enzyme study explicitly reports branched and straight-chain substrates [PMID:7698750, cached abstract; "activity toward the short branched chain acyl-CoA derivatives"]. Human patient studies independently identify this particular metabolic block [PMID:10832746, cached abstract; "2-methylbutyryl-CoA dehydrogenase"; PMID:11013134, abstract and accessible methods/figures, PubMed]. ACADSB itself performs the dehydrogenation, so its existing isoleucine process annotation is supported by participation as well as perturbation. The reaction is one step of the pathway; the earlier unsupported wording calling it the committed step was removed.
Two ontology distinctions materially change the earlier interpretation:
The butyryl evidence is assay-dependent. PMID:7698750 reports recombinant human activity. The accessible figure 3 caption of PMID:11013134 reports no butyryl or isovaleryl activity above the control background in the transfected COS-7 experiment. That paper's methods used cleared cell lysates, added FAD, ferricenium acceptor and HPLC product detection. Neither result alone justifies a universal assertion about all assay systems. The two fatty-acid metabolic process annotations remain non-core: the positive substrate chemistry supports retention, while the papers examined do not establish the magnitude of ACADSB-dependent whole-cell fatty-acid flux. The null COS-7 result is not treated as disproving every earlier activity result or all physiological contribution.
The 2011 study directly assayed purified human SBCAD with 2-methylbutyryl-CoA, in addition to characterizing valproyl-CoA turnover and inhibition [PMID:21430231, PubMed figures 3–5; cached introduction/discussion]. It supports the catalytic core. Its discussion explicitly leaves clinical toxicity causation unresolved; no toxicity, drug-response or new xenobiotic process annotation is proposed. The original 2003 substrate-specificity paper [PMID:12855692, PubMed abstract] reports differences between rat and human substrate preference despite high sequence identity. Its full paper/cache was not recovered, so exact kinetic tables, catalytic efficiencies and residue-level mechanistic conclusions are not incorporated into the review. In particular, Km alone is not used to rank catalytic efficiency, and the rat-inferred human UniProt (2R)-substrate reaction is not promoted into an experimentally tested human core.
PMID:11013134 figure 4 details import of in-vitro-translated human precursor into isolated rat mitochondria, processing and native PAGE assembly. This directly supports human-protein mitochondrial targeting and tetramers in that experimental setting. The accessible immunostaining-method paragraph describes ACAD8, not an additional ACADSB microscopy experiment. The mitochondrial-matrix annotation also relies on curated UniProt and Reactome compartment knowledge; organelle import is not described as microscopy resolving the matrix.
The primary human 2JIF structure has a D2 homotetramer biological assembly, with FAD modeled for each chain. This independently corroborates cofactor binding and oligomerization. The identical-protein-binding row is kept as a non-core positive self-association observation. FAD binding is part of the same dehydrogenase mechanism, not a second independent core function. The construct truncation discussed by Reactome is not used to infer a physiological targeting-peptide cleavage position.
General mitochondrial location, acyl-CoA dehydrogenase activity, CH-CH oxidoreductase activity and carboxylic-acid catabolism are core-compatible annotations even when more specific descendants are present. Their breadth alone does not warrant non-core or over-annotated status. Nine previous NON_CORE decisions and one previous OVER decision were changed to ACCEPT on that basis, with source-specific biological support.
All 13 IBA/IEA rows now carry propagation_review with traced source_entities. The IBA sources are ancestral nodes PTN000744030 and PTN000097586. The target's own experimental evidence among descendants is legitimate, not circularity; no independent reconstruction of the full tree/MSA is claimed. InterPro matches support broad catalytic/cofactor architecture, not arbitrary family-wide substrate transfers. Rhea/EC entries were read in the human UniProt catalytic records, distinguishing stereospecific 48256, unspecified 43780, isobutyryl 44180, butanoyl 24004, generic short-chain 47196 and hexanoyl 43464. ARBA rule identifiers are traced but detailed predicates were not inspected, so their source statuses remain UNRESOLVED even where independent human biology supports ACCEPT or NON_CORE. SL-0170 maps the explicit curated matrix statement; UniPathway UPA00364 maps the isoleucine-degradation role.
No new annotation was needed. The cached GO-CAM index has no ACADSB/P45954 hit; no missing model was interpreted as a curation gap. The two prior additional core entries were consolidated into the single enzyme mechanism. Disease context is limited to the established biallelic biochemical defect; the earlier uncited epidemiological and generalized benignity statements are not carried into the standalone description.
This follow-up addresses the review of head 289f06cbb4e4c84ae5a08e8bdb5d0eeb57faa300. The five local gene files matched the GitHub contents API at that exact head before editing (review blob 0acf9be4e5a9c3940300b098a910976ccd362bd0). All 28 seeded annotation objects, both alternative products, the single integrated core and the 24 ACCEPT / four KEEP_AS_NON_CORE decisions are preserved. Only reviewer-authored narrative and supporting evidence are refined.
The description now distinguishes straight-chain butyryl-CoA and hexanoyl-CoA from branched substrates. It retains the 2-methylbutyryl-CoA dehydrogenase synonym, mitochondrial processing, matrix homotetramer, ETF-linked isoleucine step and C5 2-methylbutyrylcarnitine accumulation. Broad expression with highest levels in heart, muscle and liver comes from the cached UniProt TISSUE SPECIFICITY record. The substrate-assay and physiological-scope cautions belong here and in the annotation rationales, rather than in the gene's standalone description. The additional substrate activities do not imply an equivalent physiological role for each substrate.
The GO-owned AmiGO term pages were inspected on 2026-09-26:
The broad catalytic ACCEPT rows retain valid source-level coverage of the same experimentally supported catalytic mechanism. For PAINT, this preserves the assessed ancestral node without claiming that its evidence resolves a single substrate. For the broad electronic catalytic parents, the independent human enzyme data establish the chemistry and multiple substrates. These are not extra core-function entries or proposed NEW ancestors; the synthesized core remains the specific isoleucine reaction. In contrast, KEEP_AS_NON_CORE for the C6 activity records a positive additional substrate reaction whose central physiological role is not established. The same distinction explains the secondary fatty-acid process rows. Term breadth alone is not the non-core rationale.
The suggested sibling comparison was made read-only: the existing human IVD review marks GO:0070991 OVER on its hexanoyl-CoA reaction, arguing that a dedicated medium-chain activity implies MCAD. That enzyme-name argument does not follow from the GO definition. It is therefore not a sound reason to reverse the ACADSB decision supported by the human substrate report. No IVD source reassessment or change is made in this gene-specific follow-up.
The complete original article was read through the University of Amsterdam repository PDF, DOI 10.1086/303105. Methods on journal page 1096 describe human constructs expressed in COS-7 cells, cleared lysates, ferricenium electron acceptor and HPLC product detection. Figure 3 on page 1100 (PDF page 8, including repository cover sheets) states: "None of the transfected cells showed activity above background when either butyryl-CoA or isovaleryl-CoA was used as substrate (not shown)." This is an assay-specific observation, not evidence that every earlier positive human substrate measurement is wrong. The import/native-PAGE experiment is separate and uses human precursors with isolated rat mitochondria. The immunostaining experiment concerns ACAD8.
The local PMID cache remains abstract-only with download-page boilerplate and has not been altered. The external recovery supersedes the earlier incomplete-access statement in these notes. PDF text was read directly; the screenshot service timed out, so no numerical bar-height inference is made. The YAML's negative COS-7 assertions were removed from action rationales and propagation comments. Those decisions now rest on the positive cached substrate report and the ontology definitions. The reference assessment gives the exact external full-text URL without pretending its figure caption is a validated cache substring. All quoted annotation supporting_text remains mechanically checkable against the unchanged caches.
The RCSB primary 2JIF record was independently rechecked: human P45954, D2 homotetramer biological assembly, FAD associated with each of chains A–D. The cofactor-binding and self-association rows now also quote the literal cached UniProt cross-reference DR PDB; 2JIF; X-ray; 2.00 A; A/B/C/D=52-432. This connects the deposited structure to the machine-owned human record; the existing cofactor and tetramer quotes remain. It does not create an additional core mechanism.
The reference-level full_text_unavailable flag remains true for PMID:11013134 because its local publication cache is incomplete. This cache-access flag does not negate the externally recovered original article described above.