Reviewed human cyclin D2 is UniProt P30279. The source contains 77 GO assertions, 36 reference records and two named protein products. All original term IDs, evidence codes, references, qualifiers, ordered supporting entities and product objects are preserved. The source is not a kinase merely because its complexes phosphorylate proteins: CDK4 and CDK6 provide the catalytic activity, while CCND2 provides cyclin regulatory activity. The frozen ClinGen gene–disease association and its classification are unchanged.
The manual assessment covers every assertion and reference. The 23 publication abstracts and seven complete Reactome summaries were read. True full-text flags are preserved as source metadata, not treated as proof that every target assay or supplement was independently examined. The detailed access ledger distinguishes complete primary narrative reading, selected complete body passages, abstract-only sources, partial HTML and uninspected target tables.
Normal deep research was attempted in an isolated temporary gene directory. The first normal wrapper attempt failed while initializing the restricted default uv cache. A second attempt with a writable temporary uv cache failed at the restricted uv tools directory. Neither attempt reached the research provider or produced a provider report. Their original execution envelopes are retained in the scientific proposal. These are manual research notes; no provider-named research file was fabricated.
PMID:8114739 explicitly identifies human CDK6, association with D1/D2/D3 in human-cell lysates, and activation following cyclin coexpression in Sf9 cells. Its canonical cache remains abstract-only; the nominal Full Text section is only an abstract and navigation material. The source supports the regulatory cyclin/CDK distinction without a claim that an isolated cyclin phosphorylates RB.
The official publisher PDF for PMID:18827403 was recovered at https://www.jstage.jst.go.jp/article/csf/33/2/33_08019/_pdf after browser HTML/PDF attempts failed. Complete Methods, Results, Discussion and main captions were read from its text extraction. This separate external reading does not change the canonical abstract-only cache or authorize quotations that are absent from that cache. No figure pixels were reanalysed.
The primary Methods specify that the full-length CCND2, CDK4 and other constructs were amplified from human HEK293T cDNA. The study uses tagged proteins in human HeLa and HEK293T cells, and separate human-construct experiments in mouse NIH-3T3 cells. Yeast two-hybrid and HEK293T coimmunoprecipitation compare intact CCND2 with an engineered N-terminal deletion. Cell counting includes HeLa and NIH-3T3 lines. Detailed synchronized thymidine incorporation, DNA-content measurements and RB phosphorylation are in NIH-3T3 cells. Those assays are not relabelled as endogenous human-cell measurements. The engineered deletion is not assumed to equal the separately named natural truncated CCND2 product.
PMID:11896535 provides independent human B-CLL evidence: stimulation increases D2/D3, functional D2-CDK4 immune complexes phosphorylate RB in vitro, and CDK6 association is not observed in those particular B-CLL conditions. This conditional result is compatible with CDK6 association in other cells and is not generalized as a universal exclusion.
The core uses the existing CDK serine/threonine kinase activator MF, G1/S process, cytosol/nucleus and D2-CDK4 complex. Broader regulator/activator ancestors remain supported source assertions but are not repeated as additional core functions. There are no NEW assertions. The production GO-CAM 61e0e55600000624 already represents P30279 as an activator in G1/S and separately as a substrate of AMBRA1-directed degradation. It does not support assigning CCND2 a ubiquitin-ligase activity merely because CCND2 is degraded.
The full PMID:18827403 paper supports the chromatin and nucleolar assignments beyond its abstract. Table II reports the tagged human CCND2-CDK4 localization populations in HeLa cells; the Results discuss heterochromatin-associated localization, nuclear-envelope arrest and nuclear entry. These remain non-core regional observations with coexpression limits. Nuclear-envelope association does not mean membrane insertion or a membrane structural activity. The soluble cytoplasmic/cytosolic phase and nuclear import support the broad core locations.
The HPA nucleoplasmic IDA and Reactome-derived nucleoplasm rows are retained without claiming independent reanalysis of HPA antibody images or deriving a compartment from a generic pathway title. Each cached event was read, including the D1-focused RBL2 event, RB1/RBL1 events, complex import, CDK T-loop activation and drug-binding events. GOA supplies the specific original compartment assertion; independent target nuclear evidence makes it biologically plausible. CDK phosphorylation and inhibitor binding in these events are not transferred as catalytic or inhibitor activities of CCND2.
Actual PTHR10177 PAINT data identify PTN000019791 for the broad cyclin features, PTN000808639 for positive G1/S regulation, and the non-negated MTOC IBD PTN007424001 dated 2025-12-18. The MTOC assertion is retained non-core with phylogenetic curator deference. Complete tree/MSA node placement and human MTOC imaging were not independently reconstructed, and no target-specific loss is established. The number or taxonomic spread of descendants is not used as an evidence score. P30279 appearing in a WITH/FROM list is an expected descendant experimental source, not circularity. The official MTOC definition describes the structure's microtubule nucleation/anchoring capability; it does not assign that biochemical capability to every localized cyclin.
All 34 original generic-binding rows preserve their individual partners and provenance. The current project override retains supported generic binding as non-core or refines it when the evidence supports a meaningful MF; it supersedes a general rule to remove binding merely for being uninformative. The standing instruction is documented at https://github.com/ai4curation/ai-gene-review/blob/500686f1/projects/CLINGEN_MENDELIAN.md#L2976-L2978.
Reviewed UniProt independently records the exact CDK2, CDK4, CDK5, p21 and p27 pairs. Its experiment counts are qualitative corroboration, not a substitute for each original control or a proof of direct contact in every AP-MS experiment. The primary B-CLL CDK4 row is refined to the demonstrated activator function. Other kinase-interaction rows are refined only to the officially defined protein serine/threonine kinase binding class, based on the preserved, corroborated exact kinase partner; no activation, phosphotransfer, neuronal process or atomic interface is added. This matters especially for CDK5: interaction alone cannot establish that CCND2 is its activating cyclin. The separately sourced CDK6 binding and broad activator rows are refined using their same-source explicit activation evidence.
The p21/p27 rows remain non-core physical associations. A kinase inhibitor's action is not assigned to the cyclin, and a single pair does not establish an adaptor bridging mechanism. Human binary screens, HeLa proximity ligation and AP-MS coassociation are distinguished in each reason. Target clone/isoform/variant tables remain uninspected where disclosed, and partial BioPlex 3.0 HTML is not treated as a complete target-level comparison.
The complete available Methods/Results passages in PMID:39526800 show that the Legionella-focused screen intentionally includes a human positive-reference set. Twenty of 21 expected human pairs pass its screening threshold, but the exact CCND2-CDK4 supplemental row was not inspected and is not asserted to be one of those 20. The actual human control design, curated target row and independent D2-CDK4 evidence preclude a title-only misattribution claim. Likewise, the cyclin-E2 focus of PMID:9840943 and cyclin-E/CDK2 focus of PMID:16765349 do not demonstrate that CCND2 was absent from the original assays.
PMID:33854239 was read at complete abstract and selected complete Results/Discussion/caption scope. D2 protein accumulation, increased half-life and joint D1/D2/D3 expression are distinct from the more detailed D1-centred rescue, substrate ubiquitination and CDK2 redistribution experiments. The combined human U2OS D-cyclin overexpression is not a D2-only experiment; mouse embryo and tumour data remain separate. The review uses this source for the existing broad cyclin statements and stability context, with stronger D2-specific primary support for the core mechanism.
The original PMID:24130168 abstract reports miR-29, CCND2 and MMP2 in gastric cancer. The exact CCND2 perturbation/rescue controls and the 2024 correction text could not be read. Independent official PubMed reciprocal links and the publisher establish the erratum DOI https://doi.org/10.1093/carcin/bgae070, but its corrected material and impact remain unresolved. The three source-specific proliferation, apoptosis and positive-G1/S assertions therefore remain UNDECIDED. This does not declare the paper retracted or the phenotypes false, and it does not invalidate independently supported general G1/S activity.
The candidate preserves the machine source and frozen associations while making manual review decisions. Normal validation, computed review status and rendering are performed on temporary candidate bytes before distinct whole-science review and ROOT acceptance. Repository DRAFT is computed from validation advisories; it is not caused merely by the presence of UNDECIDED decisions. Canonical application, standard history and publication remain ROOT-controlled subsequent steps.
The biological questions concern the cell-state and isoform dependence of alternative CDK partners and the conditions for an MTOC-associated CCND2 population. Source retrieval and correction-panel verification are audit tasks retained in these notes rather than presented as biological research questions.