The review retains all 15 historical source annotations and all three official product identities. The current QuickGO snapshot contains the same 15 term/evidence/reference assertions. New fetch metadata supplies eight supporting-entity backfills, retained in the evidence record without rewriting historical source dictionaries. No new annotation is proposed. The two cores distinguish nucleolar RNA association from cytoplasmic Hsp70-partner activity. GO:0042273 ribosomal large subunit biogenesis is attached to the latter core based on the combined partner-association, PA2G4 traffic/retention and rescue evidence. The 45S-association assay alone does not identify a large-subunit-specific RNA step.
Tummala et al., 2016, PMID:27346687 is the principal human study. Its protected publication cache is labelled full text/html but contains Abstract, Introduction and Discussion, omitting Methods and Results. The original PMC page was recovered separately. The following claims rely on that original, not on an assumed complete cache:
The ATPase passage in this paper describes yeast Jjj1/Ssa1 precedent and conserved HPD-motif reasoning. The recovered human Methods/Results do not measure ATP turnover. The core therefore uses Hsp70 protein binding, GO:0030544, a verified child of the previous chaperone-binding term, without adding an ATPase-activation assertion. The RNA core remains broad RNA binding, GO:0003723; native rRNA association alone does not specify the contact chemistry or all transcript targets.
Castello et al., 2012, PMID:22658674 has an abstract-only protected cache. The genuine original article with appended extended experimental procedures and original Table S1 were read separately. In Table S1, sheet HeLa mRNA interactome.txt, Excel row 674, ENSG00000168724/DNAJC21 is detected in crosslinked samples and not controls and is classified in the mRNA interactome. Its adjusted IC p-value is 0.00647067467032329, while its distinct adjusted DESeq p-value is 0.713048068629827; these statistics are not interchangeable. The row lists historical RefSeq products NP_001012339 and NP_919259.
The extended Methods, In Vivo Isolation of HeLa RBPs, describe UV crosslinking, LiCl/LiDS lysis, oligo(dT) capture and RNase release; LC-MS/MS and Protein Identification describe mass spectrometry and sequence identification. These support RNA contact in proliferating human HeLa cells, unlike ordinary native co-IP. They do not establish a particular RNA sequence, direct binding to 45S rRNA, or activity of all three present-day products. The current RNA-binding assertion is retained; the domain-derived nucleic-acid-binding assertion is refined to this same supported specificity without proposing a duplicate NEW annotation.
Huntingtin interacting proteins are genetic modifiers of neurodegeneration, PMID:17500595 was checked through the genuine original Methods and its supplementary data. Table S2 lists DNAJA5, Entrez 134218, with 23 Htt positives. Table S3, sheet 18, Excel rows 5351–5373, gives LOC134218/NM_194283.1 prey sequences against HTT 1–450 (55Q), 1–450 (23Q) baits. The study used human cDNA libraries, sequence checks and naive-cell retesting. Six positive nonexpanded human-pair IntAct records corroborate the Y2H assignment. The table contains prey sequence reads, not evidence that every current full-length isoform was tested. The paper's mouse-brain validation of selected interactions is not assigned to DNAJC21 without pair-specific evidence.
Miyamoto-Sato et al., 2010, PMID:20195357 was checked through the genuine original and Table S4/Table S5. Table S5 Excel rows 146–147 report MTERF–DNAJA5 IR_751 (440–469, class 2) and IR_752 (363–395, class 1). Table S4 rows 1100–1133 and 1135 give the sequence tags. The pair-specific pulldown, reciprocal pulldown and real-time-PCR verification cells are blank. The authors explicitly allow direct and indirect interactions in the display method. IntAct EBI-2690135 and EBI-2690148 corroborate the pair; its parent protein identifiers and separately annotated sufficient-binding regions must not be confused with a full-length assay. The primary region range and the curated sufficient-binding-region feature differ in their granularity and remain distinguishable.
Official MTERF1 is a DNA-binding transcription termination factor, but a short-fragment display association does not by itself establish functional-class recognition by DNAJC21. Official HTT likewise does not supply a demonstrated narrower activity for the screened pair. Both positive generic-binding annotations are retained as non-core under the explicit ClinGen project instruction. Neither compartment differences nor genericity is used to call the interactions false.
The actual PANTHER PTHR44029 tree places human Q5F1R6 at leaf PTN002541784 below PTN001940778. The positive cytoplasm IBD at that ancestor is dated 2026-05-30, taxon 2759, with PomBase:SPAC6B12.08, SGD:S000005171 and human Q5F1R6 among its evidence. The corresponding yeast JJJ1 and fission-yeast epr1 leaves were traced. The human target's presence among the evidence is valid experimental grounding, not circularity. The cytosolic refinement uses independent target evidence and does not allege incorrect ancestral placement.
InterPro IPR003604 is a Matrin/U1-C-like C2H2 domain and maps to nucleic-acid binding and zinc binding. Its actual target match is 311–345, overlapping the first official zinc finger. The broad nucleic-acid mapping is refined using the independent RNA experiments; no specific contact residues are assigned to that domain. Zinc binding is retained as a structural inference, without a claimed direct metal-binding or catalytic assay.
The HPA DNAJC21 record and its downloaded XML list nucleoli and cytosol, with supported reliability, in A-431, U-251MG and U2OS cells. The data record was read; image pixels were not independently reinterpreted. Current GO definitions and typed ancestry place nucleolus within nucleus and cytosol within cytoplasm. Refinements preserve the original positive broad-location observations and do not claim exclusive residence or antibody discrimination of the alternative products.
Qiu et al., 2006, PMID:16952052 is the correct DnaJ/Hsp40 family review and explicitly names DnaJC21 in the accessible abstract. The protected cache is abstract-only. Europe PMC XML retrieval failed; the PMC HTML is abstract-only and PDF requests returned an interstitial; the publisher PDF route returned an abstract landing page. The target-specific full article was therefore not read.
The historical NAS ribosome location remains UNDECIDED: the later gradient blots follow PA2G4 and RPL7, not DNAJC21, and neither precursor-rRNA association nor assembly-factor function alone settles localization in a mature ribosome or free mature subunit. The broad protein-folding assertion is retained as non-core, using the integrated human co-chaperone/assembly evidence and the definition that includes assisted multisubunit assembly. HSPA8 binding alone is insufficient, and no defined folding substrate or direct human ATPase assay is asserted. The original family-review access limitation remains explicit; it does not override the independent positive assembly evidence.
Official UniProt Q5F1R6 and independent FASTAs reconstruct 531, 576 and 544 residues. Isoform 2 inserts 45 residues at canonical Q395; isoform 3 inserts 13 residues at canonical K381. All retain the N-terminal J domain and both annotated zinc-finger regions, with insertions between the fingers. Functional equivalence and product-specific localization remain unestablished.
Chen et al., 2004, PMID:15067379 reports a human fetal-brain 531-residue clone in its accessible abstract. The genuine newly fetched cache is abstract-only; the publisher PDF is purchase-gated. UniProt maps this source to isoform 1 but records a residue-509 A/V conflict for AAP81807. Neither identical length nor shared domains identifies all experimental constructs with all present-day products.
The genuine normal fetch and publication caching succeeded. Falcon authentication failed and the configured fallback returned a quota error; no provider report was fabricated. Downloaded publications, UniProt, GOA, PAINT sources and earlier notes remain unchanged. The new notes appendix supersedes earlier unsupported ATPase and genericity wording. Primary full originals and supplementary tables were read only to the scopes stated above; no original figure pixels or unaccessed supplementary experiments are claimed.
The source extract contains literal data, not a second biological opinion: the four original workbooks' target rows and headers, eight selected MITAB records, three target/donor PAINT paths plus the positive IBD, HPA location/reliability fields, and the InterPro domain match. Workbook row numbers are one-based Excel rows. MITAB source_line is one-based; its field arrays retain the complete original records, including expansion, negative and feature fields. These data support the precise claims above without implying full-length constructs, orthogonal validation, or current-isoform mapping.
The source manifest records each original URL, capture time, byte count and SHA-256. The reproduction script accepts either public downloads or a JSON map from source filenames to supplied saved raw paths. It computes the extracts and rejects changed input hashes; it does not hardcode measurements or require credentials. Example from the bioinformatics directory:
uv run reproduce.py --input-map /path/to/saved-paths.json --check source-extracts.json
uv run reproduce.py --download /tmp/dnajc21-source-inputs --check source-extracts.json
Public APIs can change. A changed download fails the exact historical check and must not silently replace the preserved measurement. The supplied-snapshot mode was actually run successfully against all nine saved originals. Once the declared xlrd dependency is installed, the supplied-snapshot mode requires only those local files and makes no network request. This guide remains a locator and access record, not independent evidence in the YAML's supported_by lists.