CDK7: paired evidence notes

Human reference: https://www.uniprot.org/uniprotkb/P50613/entry

Primary source: PMID:15530371 The crystal structure of human CDK7 and its protein recognition properties..

CDK7, a member of the cyclin-dependent protein kinase family, regulates the
activities of other CDKs through phosphorylation on their activation segment and
hence contributes to control of the eukaryotic cell cycle. CDK7 also assists in
the regulation of transcription as part of the transcription factor TFIIH
complex. For maximum activity and stability, CDK7 requires phosphorylation,
association with cyclin H, and association with a third protein, MAT1.

The horse sequence comparison is in CDK7-bioinformatics/RESULTS.md. Transfer is assessed for this exact accession; human biochemical evidence is not equine experimental validation.

2026-10-20 — Full re-review, independent of the ProtNLM2 predictions

ProtNLM2 output was not opened or used. Deep research: just deep-research HORSE CDK7 --provider falcon failed ("Exception: Failed to authenticate" from the falcon client; no fallback succeeded), so no deep-research file exists for this folder; the review uses cached literature and the human ortholog record instead.

Identity of the benchmark accession (alternative gene model)

• A0A9L0R074 (415 aa, TrEMBL, no gene name) is one of eight UniProt models of the horse CDK7 gene (Ensembl ENSECAG00000014570). None is Swiss-Prot reviewed. K9K3U4 (346 aa, PE=2, from mRNA JL616500) is the transcript-supported product and has the same length and C-terminus as human CDK7 P50613 (346 aa).
• A0A9L0R074 and K9K3U4 are identical for residues 1-338; K9K3U4 then ends GLPKKLIF (as human), whereas A0A9L0R074 continues with a 77-residue segment (VAVKIQLSHHRHW...LLPKNWP) that has no counterpart in human CDK7 [file:HORSE/CDK7/CDK7-bioinformatics/RESULTS-rereview.md "A0A9L0R074 and K9K3U4 are identical for residues 1-338."]. It is therefore an alternative 3'-end gene model (alternative or read-through terminal exon), not a fragment.
• The kinase domain (human 12-295) is complete; K41, D137 (catalytic), D155 (DFG), S164 and the T-loop T170 are conserved [file:HORSE/CDK7/CDK7-bioinformatics/RESULTS-rereview.md "| A0A9L0R074 | 415 | 95.38 | 1-346 | 0 | 69 | K41 | D137 | D155 | S164 | T170 |"]. Only the last 8 human residues are replaced; nothing known to be required for cyclin H/MAT1 binding or catalysis is lost, but the extra segment is uncharacterised.
• New analysis: CDK7-bioinformatics/compare_models_rereview.py (inputs horse_CDK7_models-rereview.fasta; outputs RESULTS-rereview.md, results-rereview.json). Existing RESULTS.md left unchanged.

Evolutionary grounding

• PANTHER PTHR24056:SF0 (CYCLIN-DEPENDENT KINASE 7) for the horse model and for human P50613.
• PAINT node PTN000624096 (CDK7/Kin28/Mcs6 clade; seeds human CDK7, S. cerevisiae Kin28 S000002266, S. pombe Mcs6 SPBC19F8.07, fly Cdk7) carries IBD for TFIIK complex (GO:0070985), CTD kinase (GO:0008353), transcription initiation at RNAPII promoter (GO:0006367), cytoplasm and regulation of cell cycle (GO:0051726) [interpro/panther/PTHR24056/PTHR24056-paint.tsv]. Horse has no IBA rows in GOA, but the GOA IEA rows for these terms agree with the PAINT placement.
• Family review interpro/panther/PTHR24056/PTHR24056-review.yaml records horse in SF0 and says human SF0 vs SF375 is "unresolved"; the current human UniProt record lists PTHR24056:SF0, so I see no disagreement for the horse placement.

Key literature (human ortholog)

• CDK7-cyclin H is a TFIIH component PMID:7533895.
• Core-like TFIIH* lacks CAK; CAK sits in ERCC2/CAK PMID:8692841. Basis for MODIFY of the TFIIH core complex projection to TFIIH holo complex.
• Ser5/Ser7 CTD kinase PMID:26257281.
• CDK4/6-activating kinase, G1 progression PMID:23622515.

Decisions changed vs the previous pass

• GO:0008094 ATP-dependent activity, acting on DNA: REMOVE (TFIIH DNA ATPase belongs to XPB/XPD; CDK7 is a protein kinase).
• GO:0008353 CTD kinase: ACCEPT -> MODIFY to S5 (GO:0140836) and S7 (GO:0140837) kinase.
• Generic kinase terms now MODIFY to both specific activities (GO:0004693 and GO:0140836), not only GO:0004693.
• GO:0005634 nucleus: ACCEPT -> MODIFY to nucleoplasm; GO:0005524 ATP binding: ACCEPT -> KEEP_AS_NON_CORE.
• GO:0001650 fibrillar center and GO:0050821 protein stabilization: UNDECIDED -> KEEP_AS_NON_CORE; GO:0045944: ACCEPT -> KEEP_AS_NON_CORE.
• GO:0000307 kept ACCEPT: the specific CAK holoenzyme term GO:0019907 is obsolete.
• Two core functions now: CAK (GO:0004693, G1/S regulation, CDK holoenzyme) and TFIIH CTD Ser5 kinase (GO:0140836, transcription initiation, TFIIK).

Relative

Human CDK7 (P50613), ortholog in the same PANTHER subfamily SF0 and PAINT node PTN000624096. Review genes/human/CDK7/CDK7-ai-review.yaml exists but is status IN_PROGRESS with 9 UNDECIDED rows explicitly deferred ("still requires detailed assessment"), including GO:0000307, GO:2000045 and GO:0008094, whose papers are cached; it needs a full review. Modules: modules/g2_m_transition.yaml (CAK step).