DES encodes the reviewed 470-residue human desmin protein P17661. The official record has no named alternative protein products. Engineered deletions, disease variants, tagged constructs and unreviewed sequence records are not treated as functional isoforms. All 108 seeded annotation identities are preserved; the 111 raw GOA records reduce to 108 through the normal seeding rules. No new annotation is proposed.
The synthesis is one filament-forming structural activity: desmin self-associates and binds cytoskeletal partners to organize a mechanically connected network around Z discs, intercalated discs and the peripheral sarcolemma. Human cardiac myofilament measurements support its cardiac-myofibril location. These locations are nonexclusive. Purified human DES assembly in PMID 28470624 provides direct support. Its CRYAB experiments support protein-folding chaperone binding by DES: CRYAB is the chaperone, and DES is its binding partner. A text/figure-caption inconsistency prevents assignment of a precise DES binding-site boundary.
The full original of PMID 9736733 uses full-length hamster desmin with a deletion corresponding to the human patient variant. Human MCF7 and monkey COS host cells do not make that construct human. The patient observations and homologous assembly experiments support the structural interpretation with this distinction retained. The full original of PMID 9697706 is a human genetic/segregation report; it is not represented as a direct force, assembly or cardiac regulatory assay.
The originals for PMID 24200904 and PMID 26724190 could not be recovered. Their accessible abstracts report localization or assembly abnormalities, and independent human assembly and ultrastructural evidence support the existing curator assertions. No unseen Methods are reconstructed. Published clinical comments about the latter paper are not treated as a retraction or as a functional disproof. The abstract-only cache of PMID 9697706 likewise remains unchanged despite separate access to its original.
The ClinGen project instruction authorizes retaining supported generic binding as non-core when a more informative function has not been established. It does not verify uninspected pairs. Each binding decision considers the actual partner, construct, assay and current GO definition.
Current cytoskeletal protein binding describes binding to a protein component; intermediate filament binding describes binding to the assembled filament. A two-hybrid pair between DES and a keratin, neurofilament or GFAP therefore does not automatically establish assembled-filament recognition. The cytoskeletal regulatory binding refinements concern established regulatory partners such as epiplakin, myotubularin and phostensin; they do not assign the partner's regulation or catalytic activity to DES.
The interaction reconstruction uses exact human DES/partner/reference records from IntAct and MINT, supplemented by smaller primary studies where a matching database record was unavailable. Database absence is not a negative experiment. Matching DES sequence checksums identify canonical P17661, while participant features remain essential: a reference-sequence checksum does not prove that the full protein was tested.
The structured source assessments provide the per-paper methods, species and construct boundaries, original access scope and exact challenged interaction identifiers.
Three named partner products were reconstructed from official UniProt sequence features and FASTA. O75928-2 is PIAS2-alpha, with documented SUMO-ligase activity; it is not PIAS2-beta. Q9HC98-4 has an N-terminal extension and its tested 26–338 region retains the canonical kinase domain. Q9UJ78-2 is a 208-residue ZMYM5 product, so functions of the displayed 669-residue product are not assumed. The unreviewed UBE2I product Q7KZS0 is 184 residues and diverges after residue 137 from canonical P63279; its interaction is retained without automatically inheriting the full catalytic characterization of canonical UBE2I.
PMID 25394388 distinguishes human R350P muscle biopsies from mouse R349P models and uses mutation-specific antibodies. The human mutant is found in subsarcolemmal and sarcoplasmic aggregates. These observations retain their disease-variant scope. All three sarcolemma rows accept the established gene-level peripheral structural location, independently corroborated by normal skeletal-muscle localization in PMID 21135508. That does not turn the R350P aggregate experiment into normal membrane-insertion evidence.
The full original of PMID 26787680 supports transient nuclear mouse desmin in cardiac progenitors and immature cardiomyocytes. It includes desmin-null antibody controls, tagged murine constructs, confocal sections and purified nuclei. The authors distinguish this from mature cardiomyocytes, and do not resolve direct versus indirect DNA binding. Thus the human orthology annotations are retained with developmental scope, without adding a nuclear molecular core.
PMID 35959657 directly images mouse DES relative to lamin-B1-defined nuclear envelopes and N-cadherin-marked cell ends in controls and PKP2-deficient cardiomyocytes. PKP2 is the perturbation in that paper. Independent PMID 31822208 directly depletes rat desmin, observes nuclear-envelope infolding within a preserved cell frame, and tests microtubule dynamics and nesprin-3 connections. These data support structural participation in nuclear organization; they do not make DES a nuclear-envelope enzyme or establish every human cell context.
The full originals for PMID 11732910 and PMID 11694502 were not recovered. Their available abstracts explicitly support mouse desmin in an enriched fascia-adherens fraction and at the neuromuscular junction, respectively. Existing donor transfers are retained with these limits. PMID 9415431 also remains original-unavailable; the pre-existing Z-disc IDA is retained because human DES localization is independently well established, without inventing that paper's methods.
PMID 23533145 describes exosome-enriched pooled human urine proteomics, including enrichment and FDR controls and limitations from co-isolated material. The source HDA is retained without asserting active secretion or intravesicular topology. PMID 25771144 measures desmin among proteins from human cardiac myofilament preparations; its troponin perturbation results do not establish DES-specific force regulation.
All seven IBA source rows were traced through the actual PTHR45652 topology to the IBD nodes PTN000580114 or PTN002901991. Human DES is a descendant of both relevant nodes. Each cited donor identifier was located in the family tree, and the matching IBD rows were read. Target experimental evidence appearing among the descendants is legitimate grounding, not circularity. The skeletal-muscle developmental assertion is retained as a contextual role; it is not rejected for having only zebrafish donor evidence.
The current ARBA00028480 clause requiring IPR039008, IPR050405 and Primates matches the official human DES record and emits intermediate filament cytoskeleton. ARBA00026971 and ARBA00026673 emit cytoplasm and supramolecular fiber, but their complete matching target clauses were not reconstructed because some condition inputs, including FunFam identifiers, were unavailable. This is an explicit reconstruction limit, not evidence that the source mapping is false. Independent localization supports cytoplasm, and the more specific existing filament annotation covers the broad fiber term.
The Human Protein Atlas DES record reports supported ICC/IF localization to intermediate filaments, mapped to GO:0045111. This is an official record assessment, not an independent image or antibody reanalysis.
The four Reactome reaction records have typed input/output component paths through the sarcomeric complexes to the Z Disk Complex R-HSA-390588 and DES entity R-HSA-390570, explicitly human P17661 residues 1–470 in cytosol. This supports the compartment assertions. Myosin ATP hydrolysis and troponin calcium binding remain activities of the relevant proteins, not inferred DES activities.
A cache's full_text_unavailable value describes that cached record. Separately recovered genuine originals are identified above; none replaces or edits a source cache. HTTP failures, CAPTCHA pages and truncated metadata responses were not treated as full originals. Focused cache-valid quotations now anchor selected binding and localization assertions. The ten-word PMID 28470624 assembly quotation remains once in the core. Original-only excerpts from PMID 11353857 and PMID 21135508 are kept separately in DES-source-evidence.yaml, with explicit access scope, public original links and section/figure locators. No paper exceeds 25 quoted words across the final authored artifacts.
The review accepts 24 annotations, retains 39 as non-core, and modifies 45 existing assertions. In particular, cytoskeletal protein binding can refine a curated BioPlex physical association without adding a direct-contact claim; the AP-MS caveats above remain. Plectin's independently established structural class similarly permits this refinement despite the source-access limit. DDX6 is refined to ATPase binding because the current definition and official ATP-hydrolysis assignment establish that partner class. The five broad structural/process terms receive explicit, typed-path-checked replacements within the existing rows; no NEW rows are added.