Human C3 is UniProt P01024, HGNC:1318 and NCBI Gene 718. The normal source seed contains 142 assertions from 143 GOA records and 96 original references (48 PMID, 42 Reactome and six GO_REF). The single collapsed pair differs only by date: the mouse-supported GO:0097242 assertion dated 20181108 versus 20170605. Every seeded source object, qualifier, partner and original reference remains intact. There are no NOT assertions or seeded alternative products. Neither a source date duplicate nor an authored functional class is counted as a new GO assertion.
Seven raw GOA records explicitly name processed C3a, P01024-PRO_0000005910. The normal seeder does not retain that identifier as an existing-annotation field. The review therefore preserves the exact seed, documents the raw-to-seed joins, and interprets those seven rows as C3a rather than silently treating all products as equivalent. The four authored functional_isoforms entries are cleavage-product classes, not a claim that UniProt supplied four splice isoforms. C3a is residues 672–748; arginine-trimmed ASP/C3a-desArg is 672–747. Full C3b contains both the retained beta chain and the alpha-prime chain: mapping PRO_0000005911 alone would not imply that isolated alpha-prime has all full-C3b activities. C3d is mapped separately. The rodent-inferred C3 beta-c chain is not assigned a new human activity.
One ordinary Falcon research attempt, with the prescribed perplexity-lite fallback and installed client 0.2.7rc1, failed for both providers at DNS resolution. No provider output was produced. This is a manual research journal, not fabricated provider research. The saved provider assessment is explicitly filtered to exclude credential-bearing traceback content; it is not presented as an unedited raw envelope.
The user's explicit ClinGen instruction retains supported generic protein binding as non-core when no better supported activity is established. The published task policy records that instruction and its task-specific departure from the annotation-reviewer skill's informational-exclusion recommendation. It is not repository-wide policy or external maintainer sign-off. Specific complement- or receptor-binding evidence is refined where justified. Curator deference to an exact edge is disclosed and does not mean an individual screen table was inspected.
C3 does some of the work of complement activation. The complete cached abstract of PMID:6903192(https://pubmed.ncbi.nlm.nih.gov/6903192/) reports chemical modification of human C3, factor binding, convertase formation with B/D, and the thioester-dependent target-attachment model. C3b supplies its own covalent surface attachment and the nonenzymatic assembly platform; the partner proteases perform precursor cleavage. Thus the broad complement-activation term is valid across several pathways and is not simply a substrate annotation. The GZMK source PMID:39914456(https://pubmed.ncbi.nlm.nih.gov/39914456/) was checked in actual human reconstitution, C3a signaling, cell-surface deposition and relevant Methods passages. GZMK acts upstream through C2 and C4; it does not directly cleave C3. Neither GZMK, Bb nor activated-C2 protease activity is transferred to C3.
The six factor B/H/H-related binding rows are refined to GO:0001848 complement binding. The inverse activity, C3b binding by those partners, is not assigned to C3 merely because the paper describes it in that direction. The first core combines C3b complement assembly with its established opsonic contribution to phagocytic engulfment. Human C3b-IgG target experiments in PMID:3653927 and apoptotic-cell uptake in PMID:19302245 support the latter directly; the phagocyte executes engulfment while C3-derived material supplies recognition ligands. No protease or endocytic machinery activity is implied.
The C3d–CR2 core is a distinct processed-fragment receptor function. PMID:21527715 confirms the interaction while disputing the earlier PMID:11387479 structural interface. The earlier source is preserved with that finding-level limit. The human CD21/FCRL5 paper PMID:30107486 includes receptor co-engagement and calcium assays; its proposed physiological C3-IgG immune-complex arrangement is not treated as demonstrated in every configuration. No receptor activity is assigned to C3d itself.
The modern C3a experiments establish receptor ligand activity and G protein-coupled receptor signaling. Selected complete Methods and Results of PMID:37852260 were read, including recombinant human C3a, human C3AR1 expressed in insect cells, purified receptor complexes, HEK293 transducer assays and human macrophage responses. These are human constructs even when an expression host is bacterial or insect. The complete PMID:37169960 abstract distinguishes complement-anaphylatoxin recognition from chemokine-receptor recognition. Structures and receptor activation do not by themselves assay directional migration.
The official definitions matter. GO:0008009 chemokine activity specifies the chemokine family and has chemokine receptor binding as a parent. GO:0048018 receptor ligand activity describes a ligand effect on its receptor. GO:0042056 chemoattractant activity requires directed movement toward higher concentration. The modern C3a source rows are refined to receptor ligand activity, not automatically to chemoattractant activity. The broad complement-activation row explicitly on processed C3a is refined to the directly assayed GPCR pathway; the other C3/C3b complement rows remain valid.
The immutable PMID:342601 cache contains bibliographic metadata without an abstract. The complete original publisher abstract was read independently, as was the complete published correction. Human C3a was reported inactive in the tested human-PMN migration and rabbit-neutrophil skin-window conditions. The correction changes the trypsin-treated human-C3 result from active to inactive and fixes a C5a-desArg printing error. It is not a retraction. These are direct contrary source statements, not rejection from a title or lack of access. Accordingly the two positive source-linked chemokine/leukocyte-chemotaxis assertions are REMOVE. No universal absence of C3a migration effects on other cell types is asserted, and full original Methods remain unread.
The old PMID:5778786 abstract reports chemotactic activity, but the later original source fails to reproduce the relevant short-trypsin-C3/C3a activity and raises possible C5 contamination. Because neither full old assay was inspected, that source remains UNDECIDED rather than being confidently replaced by a positive chemoattractant assertion. C3AR1 ligand activity is independently established.
Three inspected GO-CAM models (67c10cc400005671, 67e5e74400000671 and 67f85f2b00004288) intentionally place C3a chemokine activity under leukocyte chemotaxis and complement activation using these exact sources. This was an explicit curator choice, not an overlooked product or missing annotation. The review flags the contrary original citation and activity definition for reconciliation while preserving all source objects. The separate human B-cell model 66187e4700002637 already represents the C3 receptor-ligand role. No NEW assertion is manufactured from model coverage.
ASP/C3a-desArg is distinguished from arginine-retaining C3a. Original metabolic observations are not blanket-refuted by the C5L2/C5AR2 binding dispute: human-plasma ASP acts on rat L6 myotubes in PMID:9059512; structure/function and triglyceride-synthesis responses are reported in PMID:10432298; and PMID:9555951 is a human postprandial correlation rather than an isolated causal storage assay. The human-C5L2-transfection paper PMID:15833747 reports phosphorylation/arrestin and metabolic responses, which do not alone establish heterotrimeric G-protein coupling. The receptor-specific binding and G-protein assertions remain distinct from retained contextual outcomes.
The official MGI C3 comparative GO table exposed actual mouse experimental provenance. Its displayed graph date is 2023, so it is not described as an exhaustive current experiment inventory. It links synapse pruning and neuron remodeling to PMID:18083105; amyloid clearance/endocytosis to PMID:20199584 and PMID:22438044; MAC activation and part of the angiogenesis evidence to PMID:15611275; and a protein-complex assertion to PMID:23487775. Human ortholog applicability is judged separately from whether the original assay was performed in human cells.
The complete cached PMID:18083105 abstract describes C3-deficient mice with failed retinogeniculate refinement and excess retinal innervation. The available main narrative of PMID:27033548 shows C3 deposition, C3-knockout protection and CR3-dependent microglial uptake in Alzheimer mouse models. These support contextual synaptic tagging; they do not show that C3 is a neuronal enzyme or that a human intervention was performed. Eye-specific patterning here means neural projection refinement, not formation of the eye globe. The cached PMID:20199584 abstract focuses on CR3/LRP blockade and amyloid-coated yeast uptake; its donor annotation is not represented as a directly inspected C3 manipulation.
The exact GOA PAINT node PTN000151285 is absent from the available PTHR11412 export. The cached complement IBD PTN002605320 is not substituted, and the alpha-macroglobulin inhibitor IBD is not transferred onto C3. Human C3 appearing among its own experimental descendants is legitimate evidence, not circularity. Both biological IBA claims have direct target support, while the exact historical node remains unresolved. The InterPro alpha-macroglobulin fold mapping alone does not settle intrinsic C3 endopeptidase inhibition; that claim is UNDECIDED, without a claim of catalytic-residue loss or invented inhibitor assay.
The distinct binding consultation covered 28 exact source rows; the C3a/ASP consultation covered 18 other rows. Their authors and source-reading limits are retained in TMP handoffs and are not presented as whole-gene peer approval. All remaining decisions and the integrated synthesis are the whole-gene owner's work. Screen rows retain exact curated partners with actual assay-framework and UniProt-edge checks, not invented supplementary-table reads. PMID:33961781 has a shorter canonical cache; selected Methods were read from an authenticated longer fetched variant, which remains in TMP. No variant-only quote is inserted into YAML.
The viral-gp13 human target assays and the particular exosome/microparticle target entries remain unverified. Primary exosome framework reading does not prove that every complement protein is a vesicle resident; one source explicitly discusses blood/other contamination. These limitations lead to source-specific UNDECIDED, not wrong-gene/species removal. Full-text availability is recorded separately from the actual passages read in every reference assessment. Figures, supplements and complete papers are not claimed inspected when only a narrative, abstract or selected paragraph was read.
The three cores use complement binding and two distinct processed-product ligand contexts. There is no added GO assertion, intrinsic complement protease, kinase, triglyceride-synthetic enzyme or invented specific complex. Ten short canonical quotation fields currently support key judgments; the aggregate budget is checked per source and no quotations are repeated in this journal. The initial 99-reference candidate passed ordinary schema, term, reference and best-practice validation. Final normal-reference integration and validation are recorded below; distinct ROOT whole-science peer remains pending. Canonical curated files remain unchanged.
The ordinary hosted reference batch recovered PMID:17068344, PMID:12899627, PMID:22438044, PMID:15611275 and PMID:23487775. ROOT exclusively imported those five exact caches after current-source checks, preserving the 93 original source paths. All five complete abstracts were independently read. Four caches remain abstract-only; PMID:22438044 contains full XML. Availability is not a claim that the full paper or supplements were inspected.
The two negative C5L2 primary abstracts now directly support the binding/G-protein uncertainty. PMID:17068344 describes radioligand and fluorescent-ligand assays reporting no C3a or C3a-desArg binding, while PMID:12899627 reports no C3a interaction and heterotrimeric-G-protein uncoupling. Original positive references retain finding-level DISPUTED assessments and their original identifiers. The metabolic response rows stay qualified non-core; uncertainty is now a genuine source conflict, not merely a missing cache. No disagreement is hidden by replacing ASP with C3a.
For PMID:22438044, selected complete Methods establish C57BL/6/C3-null/Mac-1-null mice, N9 and neonatal primary mouse microglia, synthetic human amyloid and human AD amyloid cores, uptake/siRNA assays and cortical injections. Results paragraph68 and Discussion paragraph81 additionally identify recombinant human C3a rescue of uptake in C3-null mouse microglia. The human product reagent is explicit; the recipient cells and in vivo experiments remain mouse. The article itself says aspects of the serum-free uptake mechanism are unclear. FigS9 and supplementary images were not inspected. The existing amyloid-clearance/endocytosis rows gain this direct donor support without claiming C3 is an amyloid-degrading enzyme or manufacturing a new annotation.
PMID:15611275 supplies the actual mouse laser-CNV/C3/MAC/anti-C6 donor context. PMID:23487775 supports CFHR dimerization and competition for complement fragments; its abstract does not expose the specific mouse complex assay, so that access limit remains. The earlier mouse donor table is retained as a provenance link rather than rebranded as a fully inspected current experiment census.
The final YAML preserves all 142 source objects, all 96 original identifiers/titles, 143 raw GOA records and seven explicit raw C3a product joins. It contains 69 ACCEPT, 43 KEEP_AS_NON_CORE, 18 MODIFY, 10 UNDECIDED and two REMOVE, with zero NEW assertions. The original 48 PMID references plus three already-cached contextual papers and five new normal sources yield 56 PMIDs, 42 Reactome and six GO_REF references (104 total). Three cores and four authored cleavage-product classes are unchanged from ROOT's initial whole-draft reading. Seventeen short canonical quotation fields remain within 25 aggregate words per source, with no quoted text added to these notes. The exact semantic delta from the prior f042b65e candidate is preserved in TMP.
The final normal validation passed on the 104-reference candidate with 19 warnings: 17 intentionally retained generic-binding rows under the explicit task instruction, one source/product-specific complement-activation difference, and one source-specific chemokine action difference. Status remains DRAFT; uncertainty is not converted to evidence of absence. This is a completed TMP scientific proposal awaiting distinct ROOT whole-science peer, with no canonical authored changes.
ROOT requested a citation-format refinement before final peer: the two disputed ASP findings now include exact negative-source snippets in their finding-review support, while the old chemotaxis dispute remains in the manual reference assessment and source/URL narrative because the contrary normal cache is citation-only. No external text is fabricated as a cache quotation. All actions, cores and product classes remain unchanged. The aggregate negative-source quotations total 21 words for PMID:17068344 and 18 for PMID:12899627.
The exact YAML and notes approved by the distinct internal ROOT whole-science peer were applied after preimage checks. Normal canonical validation passed with 19 explained advisories; a standard codex/gpt-6 CREATE history was scaffolded and validated, and targeted HTML rendering passed. All 142 source objects, 96 original reference identifiers/titles and seven raw C3a product joins remain intact. The final 104 references, three cores, four authored cleavage-product classes and DRAFT status are preserved. Seventeen short quote fields stay within the aggregate per-source budget. No source cache or prior history was overwritten. Remote publication is a separate ROOT-owned step.
The published review's electronic endopeptidase-inhibitor decision is revised from UNDECIDED to REMOVE. The domain matches themselves are real; the problem is transferring inhibitor specificity across functionally distinct branches of the macroglobulin family. Human C3 experiments in PMID:6903192 establish thioester-dependent attachment to a target surface and complement-convertase formation. Primary structural descriptions distinguish this complement mechanism from the protease-cleavable bait sequence and protease-trapping mechanism of macroglobulin inhibitors. The 2022 human A2ML1 structural study describes the complement anaphylatoxin segment replacing the functional bait region and distinguishes covalent attachment to a complement-activating surface from attachment to the attacking protease. The 2021 structural mechanics study likewise distinguishes human C3's anaphylatoxin/alpha-prime N-terminal segments from the inhibitor bait-region arrangement. Access was limited to the indexed primary-publisher paragraphs; full publisher opens failed, and no full-paper or figure reading is claimed. No new publication cache or quotation has been manufactured.
This is a functional-divergence argument about the electronic mapping, not a universal claim that C3 cannot influence any protease. Historical names on homologous domains, including a bait-region-domain label, do not independently establish the functional bait sequence or trapping activity. The exact GOA PAINT node remains unavailable. The earlier journal statement that the available inhibitor IBD was not transferred in this review must not be read as proof that PAINT explicitly placed that historical node outside C3. The existing IBA decisions and node-access limits are unchanged. The retracted Nature paper with DOI 10.1038/nature05258 is not used as evidence for this correction.
The two existing C3d–CR2 binding annotations from PMID:11387479 and PMID:21527715 are refined to GO:0048018 receptor ligand activity. That activity already has an accepted source row and the C3d core; the change makes the binding-row refinements consistent with the supported receptor-ligand role. The early structural interface remains disputed, and the isolated structure is not described as a measurement of every downstream immune response. The official indexed receptor-ligand term page currently displays a has-part relationship to signaling receptor binding, so this decision does not repeat the review comment's unsupported is-a assertion.
The two C3a process refinements and the corresponding single core now use GO:0002430 complement receptor mediated signaling pathway. Its official definition includes signaling triggered by complement fragments binding complement receptors. This precisely covers the human C3a–C3AR1 activation evidence in PMID:37852260 and the receptor/chimera assays in PMID:10085065. C3a is the ligand, not the receptor or GTPase. The core still describes G-protein/arrestin mechanism, but does not retain a second generic GPCR process entry. No C3aR-specific binding term was verified, so no identifier is guessed.
Other suggestions remain bounded: the normal PMID:342601 cache stays citation-only, with the already recorded original publisher abstract/correction and cell-type limits; no raw NOT field is rewritten. Surface-attachment chemistry remains explicit in the first core and biological description without inventing a GO term. Campaign tracking awaits a confirmed merge. All 142 source assertions, 104 references, four authored cleavage-product classes, three cores and existing short quotations are preserved; no NEW annotation is added. Earlier dated notes and history remain historical records.
Normal focused candidate validation passed on 2026-10-03 (actual b99bf5): 19 expected warnings remain, comprising 17 supported generic-binding advisories under the explicit task instruction and two deliberately source/product-specific action differences. The follow-up changes no quote text or reference identity. Independent science review remains pending; no canonical application has occurred.
Applied the independently reviewed inhibitor-mapping and receptor-specificity changes. Canonical validation and standard EDIT history validation passed; 19 explained warnings remain. All 142 source assertions, 104 references, four cleavage-product classes and existing quotations are preserved. DRAFT and nine unresolved source assessments remain explicit; no NEW assertion is introduced. Earlier journal entries, history and source files were preserved.