Focused Hypothesis Report — F7A4N8 (Equus caballus CTDSP2): "enables kinase activity" (GO:0016301)

Executive Judgment

Verdict: REFUTED (kinase activity not supported), with the strongest alternative being a HAD-like protein phosphatase family.

The seed hypothesis is that horse protein F7A4N8 enables kinase activity (GO:0016301). Independent
sequence and orthology analysis refutes this:

1. No kinase machinery. The exact 174-aa sequence (SHA-256 verified) contains none of the
   defining eukaryotic protein-kinase catalytic motifs (Gly-rich loop GxGxxG, VAIK β3 lysine, HRD
   catalytic loop, DFG Mg-binding, APE). It has only 3 aspartates in 174 residues — incompatible with
   an ATP-phosphotransfer active site.
2. The gene family is a phosphatase, not a kinase. CTDSP2 belongs to the FCP/SCP small
   C-terminal-domain phosphatase family (HAD-like hydrolase superfamily). Its human ortholog O14595 is a
   reviewed protein phosphatase (active-site Asp107 forms a 4-aspartylphosphate intermediate). A
   phosphatase catalyzes dephosphorylation — mechanistically the opposite of a kinase.
3. This particular horse model is truncated/aberrant. F7A4N8 matches human CTDSP2 at ~96% identity
   over only the N-terminal ~84 residues, then diverges into a Ser/Arg-rich low-complexity C-terminus.
   It lacks the entire FCP1-homology catalytic phosphatase domain (human residues 97–255, DLDET
   active site). So this sequence supports neither kinase nor (on its own) phosphatase catalysis.

Most important caveat: F7A4N8 is a TrEMBL/Ensembl automatic prediction whose C-terminus appears to be
an incomplete or mis-predicted gene model. Identity with the original prediction-time input was not
independently establishable, but the supplied sequence was verified byte-for-byte (hash match).

Reproducible Methods & Provenance

• Sequences (retrieved 2026-09-08 from UniProt REST rest.uniprot.org/uniprotkb/{acc}.json):
• Target: F7A4N8 (TrEMBL, Equus caballus), 174 aa, mass 19,012 Da; submitted name "CTD small
  phosphatase 2"; gene CTDSP2 (VGNC). SHA-256 = 94788163d6db3dbca3e1432762d7b2a1e8b57e1be2048c233c505a07a0b5042c (match confirmed).
• Human lead: O14595 (Swiss-Prot), "Carboxy-terminal domain RNA polymerase II polypeptide A small
  phosphatase 2", 271 aa; FCP1-homology domain 97–255; active sites Asp107 (4-aspartylphosphate
  intermediate) and Asp109 (proton donor); Mg binding 218; keywords Hydrolase, Protein phosphatase,
  Magnesium, Nucleus; has PDB structure.
• Motif scan (regex against the verified sequence) — kinase hallmarks and FCP/SCP DxDxT/V motif.
• Pairwise comparison F7A4N8 vs O14595 (identity of shared prefix; divergence point; catalytic-domain
  presence).
• Computed outputs are reproduced in the Evidence Matrix and Mechanistic Scope below.

Independent cross-database checks (added Iteration 2):
- UniProt F7A4N8 has no GO cross-references and the word "kinase" appears 0 times — the kinase
claim is not database-derived.
- UniProt O14595 GO annotations: GO:0008420 "RNA polymerase II CTD heptapeptide repeat phosphatase
activity" with IDA (direct assay) evidence; GO:0006470 protein dephosphorylation (IDA);
GO:0005654 nucleoplasm (TAS). "kinase" appears 0 times; "phosphatase" 13 times.
- InterPro F7A4N8 → HTTP 204 (no domain matches at all) — the horse model carries neither a kinase
domain nor the phosphatase domain.
- InterPro O14595 → FCP1 homology domain (IPR004274), Dullard phosphatase domain (IPR011948), HAD-like
superfamily (IPR036412), Pfam PF03031 "NLI interacting factor-like phosphatase", SMART SM00577
"catalytic domain of ctd-like phosphatases", SFLD "RNA Pol CTD Phosphatase Like" — all phosphatase/HAD
signatures, zero kinase signatures.

Structural check (added Iteration 3): AlphaFold DB model AF-F7A4N8-F1 (v6, 2025-03-31) — mean
pLDDT 46.2 (very low); 66% of residues pLDDT <50, 98% <70; shared N-terminus (res 1–84) mean 51.4,
divergent C-terminus (res 85–174) mean 41.4. No high-confidence globular/catalytic fold, consistent
with the InterPro no-domain result.

Key computed values:
- Kinase motifs GxGxxG / VAIK / HRD / DFG / APE: all NONE.
- FCP/SCP catalytic motif DxDx[TV] / DLDET in F7A4N8: NONE ("DLDET" absent).
- F7A4N8 vs O14595 identity over first 84 aa: 81/84 = 96.4%; sequences track together until human
~res 84 (…CLQYQFYQ), then diverge. Human DLDET catalytic motif at residues 107–111 has no counterpart
in the horse model.
- F7A4N8 composition: Ser 16.7%, Arg 8.0%, Asp 1.7% (Asp at positions 14, 73, 147 only).

Evidence Matrix

GO Curation Implications

• GO:0016301 kinase activity — REMOVE / do not annotate. No sequence, structural, or family evidence
  supports kinase activity for F7A4N8 or the CTDSP2 family. This is a misannotation (a phosphatase gene
  annotated with the mechanistically opposite activity). (Lead — requires curator verification.)
• Better-supported MF (by orthology/family, requires the catalytic domain):
  GO:0004721 phosphoprotein phosphatase activity (or more specific
  GO:0008420 RNA polymerase II CTD heptapeptide repeat phosphatase activity /
  GO:0004725-adjacent serine/threonine phosphatase), plus GO:0046872 metal ion (Mg) binding.
  Caveat: these should be transferred to the horse gene with an ISS/orthology qualifier and a note that
  the current F7A4N8 protein model lacks the catalytic domain — the annotation belongs to the gene, not
  necessarily to this truncated protein isoform.
• BP/CC (family-level, orthology): BP GO:0006470 protein dephosphorylation /
  transcription regulation; CC GO:0005634 nucleus.
• Avoid "protein binding" as a terminal recommendation — the phosphatase MF is more informative and
  well-supported for the gene.

Mechanistic Scope

• Immediate molecular function tested: ATP-dependent phosphotransfer (kinase). Not supported — no
  ATP-binding/kinase fold in the sequence.
• Established family function (direct gene-product activity): Mg²⁺-dependent HAD-family
  aspartate-nucleophile phosphatase that removes phosphate from phospho-Ser substrates (RNA Pol II CTD
  Ser5; REST degron phosphosites). This is a hydrolase reaction, not phosphotransfer.
• Downstream/phenotypic (not the MF): transcriptional regulation, REST stabilization, neuronal gene
  silencing, neurogenesis timing — consequences of phosphatase activity, not the activity itself.

Conflicts and Alternatives

• Reaction-direction confusion (kinase vs phosphatase): The most likely source of the erroneous
  "kinase activity" term. Both terms concern phosphate, but the gene is a phosphatase. No paralog of CTDSP2
  is a kinase.
• Truncated/aberrant gene model: F7A4N8's C-terminus (res 85–174) is Ser/Arg-rich low-complexity and
  diverges from all mammalian CTDSP2 orthologs, which continue into a conserved catalytic domain. This
  looks like an incomplete or frameshifted Ensembl prediction rather than a genuine catalytic-domain-less
  isoform; it warrants caution before transferring any enzymatic MF to this specific protein record.
• Isoform/orthology: Identity is confirmed as horse CTDSP2 by gene name and VGNC/Ensembl xrefs and by
  the 96% N-terminal match to human CTDSP2 — but the protein model is not full-length.

Knowledge Gaps

Discriminating Tests

1. Retrieve full-length horse CTDSP2 (Ensembl/RefSeq) and re-scan for the DLDET motif and FCP1 domain
   — expected present in the true gene, absent in F7A4N8's current model.
2. HMM/domain search (Pfam PF03031 "NIF"/FCP1-like) and structure (AlphaFold F7A4N8) to confirm the
   catalytic core is absent from this model.
3. Phosphatase activity assay (para-nitrophenyl phosphate or phospho-CTD peptide) on recombinant
   full-length horse CTDSP2 — positive would confirm phosphatase; a kinase assay (ATP transfer) would be
   negative.

Curation Leads (require curator verification)

• Action: Remove/reject GO:0016301 "kinase activity" for CTDSP2 (F7A4N8) as a misannotation.
• Replacement MF leads: GO:0004721 phosphoprotein phosphatase activity /
  GO:0008420 RNA Pol II CTD phosphatase activity; add GO:0046872 magnesium ion binding; BP
  GO:0006470 protein dephosphorylation; CC GO:0005634 nucleus — all as orthology (ISS) with a note
  that the current F7A4N8 model lacks the catalytic domain.
• Candidate references to verify: PMID 30217818 (SCP1 phosphatase kinetics + structure), PMID 17403776
  (SCP1 phosphatase in CNS development), PMID 38609948 (SCP1 as a phosphatase drug target); database:
  UniProt O14595 (human CTDSP2, reviewed phosphatase).
• Suggested question for curator: Should enzymatic MF terms be attached to the F7A4N8 protein record at
  all, given it is a truncated model missing the catalytic domain — or annotated at the gene level only?

Limitations

• No directly observed Equus caballus biochemical data were found; family/ortholog transfer relies on
  human/mouse SCP paralogs. Motif and alignment analyses are computational. PubMed queries for "SCP2/CTDSP2"
  specifically returned no results; cited evidence is for the closely related paralog SCP1 and the family.

Artifacts

• OpenScientist final report
• OpenScientist final report