The existing review contained 82 source records and four repeated DNA methyltransferase core entries. This review preserves those source dictionaries, adds 22 genuinely distinct normally seeded records and assigns a verdict to all 104 records. The final distribution is 30 ACCEPT, 29 KEEP_AS_NON_CORE, 37 MODIFY, 6 UNDECIDED and 2 MARK_AS_OVER_ANNOTATED. No new GO assertion is proposed. The provisional status is DRAFT; ordinary validation determines the final status.
The historical GOA snapshot is preserved byte for byte. The current GOA is the unmodified normal-fetch result, not a concatenation or hand-edited TSV. It contains 87 rows and 86 unique term/evidence/reference/partner tuples. The review keeps five genuine new partner splits and 17 other identities. Eight additional normal-seeding records only repeated legacy claims with updated propagation metadata and were not appended. Eighteen historical identities receive retired: true because they are absent from the new snapshot; their original fields and scientific verdicts remain visible. Retirement does not mean that the biological assertion is false. The exact bridge, hashes and historical execution provenance are in DNMT1-source-provenance.yaml.
All preexisting UniProt, publication, Reactome, PAINT, provider, pathway and visualization bytes remain protected. The old manual and Falcon research files are retained. A fresh configured Falcon attempt failed authentication and its genuine Perplexity-lite fallback failed quota authorization; no replacement provider report was authored. The authored manual research citation was removed as an independent support source, not because its file was absent.
Direct human substrate assays support the CpG-specific methyltransferase activity GO:0051718. Its typed ancestry includes the broader original cytosine-5 methyltransferase term. The historical fixed twofold preference was removed: original kinetics are construct- and condition-dependent (2011 structure/kinetics, 2015 human structure). The four repeated catalytic blocks are consolidated into one maintenance-methylation unit. Two complementary H3K18ub and H3K23ub reader entries represent the same cooperative H3Ub2 module, not two independent methylation pathways or full activation by either isolated mark (2022 activation study). The inaccessible 2017 full original limits the specific H3K14 claim, which remains UNDECIDED.
CpG-island silencing is anchored to actual human promoter methylation/expression experiments (USP7/DNMT1 study, KRAS/ZNF304 study). The heterochromatin role also has human loss-of-function and mouse-rescue support; its catalytic contribution is inferred alongside possible scaffolding effects, not isolated experimentally (original article). Maintenance methylation, CpG-island silencing and constitutive heterochromatin formation are not ancestor/descendant duplicates in the checked typed ontology closure.
The explicit ClinGen project instruction authorizes retention of supported generic interactions. Positive native complexes and affinity-capture pairs remain non-core where a specific recognition activity is unestablished. Verified binary/fragment binding plus the actual partner class supports narrower UHRF1, NRIP1, SETD7, AKT1 and USP7 terms. These refinements do not confer the partner's catalytic activity on DNMT1. The unresolved viral-screen pair and UBE2I pair remain UNDECIDED rather than being called false positives.
Mouse Kcnq1ot1 and Dum experiments support lncRNA association through native RIP and extract pull-down; neither demonstrates purified direct RNA contact. Methyl-CpG binding includes the directly assayed hemimethylated substrate and does not require fully symmetric methylation. The two diet/BPA response transfers are over-annotated: their observed expression response is real, but the studies do not establish DNMT1 as an executing response factor.
Nuclear annotations are refined to a supported nucleoplasmic pool using human HPA immunofluorescence and explicit Reactome protein entities. Each original source's broader resolution is retained in its reason; the refinement is nonexclusive. Chromatin is also explicitly represented in all three core units: nucleoplasm excludes chromosomes by definition, whereas the maintenance and H3-reader functions operate on chromatin. The two compartments are not typed ancestor/descendant duplicates. The MitoCoP protein-group evidence supports a contextual mitochondrial pool, without assigning mitochondrial DNA methylation or one natural isoform. The portable source extraction preserves the literal MitoCoP and OpenCell records.
The official products have 1616, 1632 and 1280 residues. Product 2 replaces one canonical residue with a 17-residue segment; product 3 lacks residues 1–336. Human structural fragments beginning at residue 351 are engineered constructs and are not natural product 3. Retained domains do not prove equal activity or recruitment of all three products. Source evidence and access limits distinguish complete originals, selected original sections, abstract-only caches and separately recovered data. Six short new exact anchors are used once each in the review; reference-only support avoids repeating excerpts. Earlier title-only snippets are superseded by scoped source findings, while all cached source text remains unchanged.
The six broad nuclear annotations retain their source assertions and MODIFY actions, but now preserve both the supported nucleoplasmic and chromatin-associated pools. Rows 1, 7, 63, 81 and 89 use nucleoplasm plus chromatin. The source-specific row 90 uses nucleoplasm plus pericentric heterochromatin: PMID:24492612(https://pubmed.ncbi.nlm.nih.gov/24492612/) Results and Figure 4B caption explicitly identify endogenous human DNMT1, PHF20L1 and 5-methylcytosine in pericentric heterochromatin in HeLa cells. That term already lies beneath chromatin, so the generic parent is not redundantly listed on the same row. These are nonexclusive pools, not assertions that every natural product has identical localization.
The HPA record provides supported nucleoplasm ICC/IF data, not evidence against chromosome association. No independent image-pixel interpretation is claimed. PMID:12145218(https://pubmed.ncbi.nlm.nih.gov/12145218/) distinguishes endogenous human nuclear extracts from human DNMT1 residues 1–446 expressed in COS-7. The historical screen/TAS sources remain limited to their actually accessible abstracts; finer gene-level locations are supported independently, not attributed to an unseen target table. Replication-fork row 35 is retained separately. Toroidal nuclear signal or a cell-cycle context alone was not promoted to a new fork-localization assay in the six nuclear rows.
Current chromatin is part of the live chromosome term, not an obsolete synonym. Neither chromatin nor pericentric heterochromatin has a typed nucleus ancestor in the inspected is_a/part_of graph; the biological nuclear pools are established by experiments. The chromosome-to-chromatin reasons now state the actual specificity relation. The broad transcription refinement likewise identifies the performed CpG-methylation regulatory mechanism without attributing polymerase catalysis to an upstream association.
The two H3K18ub/H3K23ub reader entries remain complementary representations of one cooperative module, as their unchanged descriptions state. PMID:36414620(https://doi.org/10.1038/s41467-022-34779-4) separates human structural/biochemical and cell-rescue experiments from Xenopus chromatin assays; those scopes are not interchanged. Both supported reader terms, the catalytic core, product identities, all 104 annotation source objects/actions, all reference assessments and six inherited supporting quotations are unchanged. No NEW assertion or newly reproduced publication quote is added.
The public source-provenance command and working directory are normalized to portable repository-relative form while retaining the actual fetch timestamps, exit code, source hashes and historical/current lifecycle bridge. This does not replace or alter the authentic execution receipt. The audited analysis remains fail-closed on source-hash mismatch. Prior notes and history are unchanged; any new history is append-only.