ERCC5/XPG is best characterized as a nuclear DNA endonuclease that recognizes DNA junctions and performs the 3′ incision in nucleotide-excision repair. Human reconstitution, purified-protein incision, footprinting and current pathway catalyst records agree on this role. The human protein is not assigned the helicase chemistry of its TFIIH partners. Additional experiments support repair-protein assembly/loading in homologous recombination and stimulation of NTHL1 in biochemical base-excision repair assays. These additional roles remain distinct from XPG nuclease catalysis. PMID:11259578(https://pmc.ncbi.nlm.nih.gov/articles/PMC86862/), PMID:12644470(https://doi.org/10.1074/jbc.M213155200), PMID:26833090(https://pmc.ncbi.nlm.nih.gov/articles/PMC4761302/), PMID:9927729(https://academic.oup.com/nar/article/27/4/979/2902566).
All 77 imported annotation contexts have been reviewed: 35 ACCEPT, 14 KEEP_AS_NON_CORE, 20 MODIFY and 8 UNDECIDED. No imported source object or natural-product identifier was changed. No NEW annotation or publication quotation was added. Core-function synthesis remains a separate next stage. The source guide records actual original access, operative locators and specific unread material; availability flags in the normal publication caches remain untouched.
Full-length XPG and engineered reagents must be kept separate. The XPGcat structural reagent replaces a large segment with archaeal FEN sequence and has a different substrate range from full-length XPG. Another structural study links human residues 1–112 to 750–990. Neither is the natural 419-aa or 232-aa splice product, and the engineered ΔN136 reagent used in earlier self-association experiments is not a natural isoform. The functions and localization of the two shorter natural products remain unresolved. PMID:32522879(https://pmc.ncbi.nlm.nih.gov/articles/PMC7321962/), PMID:32821917(https://pmc.ncbi.nlm.nih.gov/articles/PMC7515719/), PMID:10026181(https://doi.org/10.1074/jbc.274.9.5637).
Human XPG contacts both the duplex and a short single-stranded segment beside a junction. This is not a claim of unrestricted high-affinity binding to isolated DNA strands. Isolated circular ssDNA cleavage is independently demonstrated in the original human study, rather than inferred from bubble-DNA incision. The specific damaged-DNA binding assertion remains uncertain because the inspected structural substrate is synthetic splayed-arm DNA; the GO term itself does not require preferential affinity for damaged DNA. PMID:12644470(https://doi.org/10.1074/jbc.M213155200), PMID:8078765(https://pmc.ncbi.nlm.nih.gov/articles/PMC523723/?page=0).
Nucleus and chromosome terms were refined to nucleoplasm and chromatin where controlled human repair-site evidence provides that specificity. The HPA record has an explicit cross-gene antibody caution. The biochemical P100 fraction contains chromatin, nuclear matrix and other insoluble material; it is not purified chromatin. These limits remain in the review. The electronic source mappings are valid, and the refinements address term scope rather than claiming a source misattribution. Human localization study, HPA ERCC5.
The TC-NER annotation is supported by an explicit human XPG catalyst chain: pathway R-HSA-6781827 contains dual incision and the XPG 3′-incision reaction, whose active unit maps to full-length human P28715. This establishes participation through a defined step. It does not verify all upstream recognition or polymerase-backtracking models, nor substitute for access to the original PMID:16246722 interaction assays. TC-NER pathway, XPG incision.
The eight remaining uncertainties concern the precise RNA polymerase II, CSB, p62 and p44 interaction claims; the damaged-DNA binding claim; the inaccessible ERCC5–BRAF supplementary screening record; and the original RPA-associated claim. RPA association is not recast as constitutive RPA-subunit membership. The review preserves positive curator assertions as undecided when their precise experiments cannot be assessed.
Supported generic binding is retained as non-core under the standing campaign instruction unless an evidence-backed specific role is available. NTHL1 activation and verified TFIIH partner classes are therefore refined, while other supported interactions are retained. The no-new-quotation instruction also applies. These choices preserve the measured biology without adding an untested adaptor, enzyme or process role.
Automated deep-research attempts did not yield a usable provider report. The accompanying manual research synthesis is explicitly manual. See the source guide for direct original URLs, figure/section locators, database provenance and actual access limits. This annotation checkpoint passes schema and preservation checks; final core, computed-status, full validation and rendering checks follow in the finalization stage.
One core unit now represents DNA endonuclease activity in nucleotide-excision repair. Human full-length XPG cuts structured DNA junctions and covalently closed ssDNA; the selected catalytic term covers both measured substrate geometries. Its principal repair action is the 3′ incision in global-genome and transcription-coupled NER. The NER process term covers both without a redundant parent-and-child pair. Human ssDNA nuclease experiments, full-length junction assays.
The complex context describes functional repair assemblies, not permanent preassembly or constitutive transcriptional TFIIH membership. Human reconstitution and salt-sensitive associations support this distinction. Current human GG-NER and TC-NER catalyst records each identify full-length P28715 as the DNA-endonuclease active unit and a direct member of the respective incision assembly. The noncatalytic HR recruitment and NTHL1 activation roles remain separate non-core functions; neither is assigned to XPG nuclease chemistry. All 77 annotation reviews, all reference assessments and the three natural-product records are unchanged. No new annotation or publication quotation was added. Functional human repair complexes, GG-NER incision, TC-NER incision.
Normal whole-gene validation, strict whole-document term validation, component checks, computed-status checking and rendering all passed on the final review. All 156 description, annotation-review and core narratives occur in the rendered page. Status is DRAFT with eight unresolved source-specific assessments and ten advisories: four supported generic-binding rows retained as non-core under the campaign instruction, five accepted rows without structured supported_by entries, and one unused manual-research reference. The original reference identifiers, explicit reasons and source guide retain the evidence for those five rows; no publication quotation was added merely to remove an advisory. No warning suppression or automatic NEW annotation was used.
All 77 imported source objects and annotation decisions, 41 reference entries and three natural products remain unchanged from the accepted annotation/core synthesis. The genuine GOA and UniProt files and all normal source caches remain exact. The selected dependencies comprise 24 cited publication caches and 11 Reactome caches. Uncited fetch-only papers and PANTHER acquisition auxiliaries remain outside the public changes. Automated research-provider failure is documented as such; the biological research synthesis is explicitly manual.
Canonical publication checks also passed: normal validation, strict whole-document term validation, ontology-aware component checks, computed status and rendering. All 156 description/review/core narratives occur in the generated page. DRAFT retains the same ten documented advisories; every selected normal-source file remains byte-identical.
The DNA-binding replacement set now includes all three supported descendants, with junction-contact limits retained. The generic complex category is TOO_GENERAL while the HR assembly evidence remains recorded. NEF3 retains phylogenetic and human repair-assembly support with the measured 150 versus 300 mM KCl stability limit. The damaged-DNA claim remains undecided because the supplement and figure pixels were not inspected; no requirement for preferential lesion affinity is imposed. The NTHL1 rationale now explicitly records the positive FLAG-XPG stimulation in the 2004 abstract. Five accepted rows gain reference-only structured support, and one proposed experiment addresses natural-product expression and function. Current official ancestry shows GO:0001097 and GO:0062058 are not an ancestor/descendant pair, so their distinct partner scopes remain unchanged. No source, reference record, product, core or publication quotation changes.
Canonical follow-up validation passed: normal, strict whole-document terms, components, computed status and rendering. DRAFT retains 5 advisories. All 156 description/review/core narratives occur in rendered HTML. All 35 selected normal-source dependencies remain exact.