TRMT10C notes

Literature review was done manually from cached PubMed records; the falcon deep-research report (TRMT10C-deep-research-falcon.md) completed afterwards and was skimmed for consistency with the review decisions; no conflicts were found.

• MRPP1-MRPP2 subcomplex is the m1G9/m1A9 methyltransferase PMID:23042678.
• Mitochondrial RNase P is a protein-only enzyme from three proteins PMID:18984158.
• MRPP1/2 acts as a tRNA maturation platform for ELAC2 and TRNT1 PMID:29040705.
• Disease (COXPD30): variants destabilise MRPP1 and impair processing without abolishing methylation PMID:27132592.
• Nucleus/nucleoplasm IBA come from a PANTHER node above the TRMT10A/TRMT10C split (donor TRMT10A, Q8TBZ6); removed.

Comparator check behind the proposed contributes_to GO:0004526

CLAUDE.md requires a comparator check before a NEW, naming two or three gene products
standing in the same role. The role here is "non-catalytic protein subunit of an RNase P
holoenzyme", and the nuclear holoenzyme supplies ten of them: its catalytic moiety is the
RPPH1 RNA, not any protein, so GO must say something about the proteins' relation to the
activity. Query run 2026-10-12 against QuickGO:

curl -H 'Accept: application/json' \
  'https://www.ebi.ac.uk/QuickGO/services/annotation/search?goId=GO:0004526&taxonId=9606&limit=100&goUsage=exact'

40 human annotations. Grouped by symbol and qualifier:

GO:0005655 is nucleolar ribonuclease P complex, GO:0030677 ribonuclease P complex and
GO:0030678 mitochondrial ribonuclease P complex. Their ancestry is read from runoak relationships on 2026-10-12 rather than from memory, and it is not the flat parent/child shape
an earlier draft of this paragraph asserted: GO:0030677 is the direct is_a parent of
GO:0030678, but reaches GO:0005655 only through GO:0030681 multimeric ribonuclease P
complex, so it sits two levels above the nucleolar term and one above the mitochondrial one.
The asymmetry does not change the reading below, since GO:0030677 is still the term both
specific complexes descend from, but "generic parent" was wrong for the nucleolar side.

The column records the term each gene product is actually annotated to rather than a
nuclear/mitochondrial reading of it, which is a third query and is logged below. It does not
apply to the two ComplexPortal rows, which are the assembled complexes themselves rather than
parts of one; those two cells read n/a, is the complex and come from no query. Two things the
column shows that "nuclear" would have hidden: every one of the ten protein subunits is in the
nucleolar term, while none of the fourteen ncRNA ids is - they carry GO:0030677 and
nothing below it, RPPH1's four included. So the ncRNA enables rows are not recorded as
nucleolar-complex members at all, which is consistent with reading them as a Rfam-style
electronic propagation rather than as curated subunit annotations.

Rows sum to 40, which is the hit count, so nothing was truncated at limit=100.

The three bold rows are the exceptions, and they are in the table rather than only in the
prose below for the same reason the ComplexPortal rows are: a reader who checks the table
and stops there would otherwise come away with the absolute claim this file spent a round
retracting. The two ComplexPortal rows are assembled complexes rather than proteins, which
is why they carry enables and why the claims below are scoped to proteins; CPX-6155 is
the only enables row with a direct-assay code other than PRORP's.

Note the complex column: PRORP (O15091) is part_of GO:0030678, the mitochondrial complex,
not the nuclear one. The nuclear catalyst is an RNA, so the enables rows there sit on ncRNA entries
rather than on any protein subunit - four of them RPPH1 itself and ten further RNAcentral
ids that are not named as RPPH1 - but not absolutely: POP5, RPP30 and RPP40 each also carry an
enables row, POP5 by IEA and the other two by TAS. What holds without exception is the
evidence-code version. Every IDA row on a nuclear protein subunit is contributes_to, and
PRORP is the only protein in either complex carrying enables by direct assay, because the
mitochondrial catalyst is a protein. The experimental curation lines up behind contributes_to
for the subunits; only the electronic and author-statement rows do not, which is the propagation
class this repository treats as arguable.

So GO's convention for an RNase P holoenzyme is enables for whatever does the chemistry -
the RPPH1 RNA in the nucleus, PRORP in mitochondria - and contributes_to for the other
protein subunits. That is exactly the shape of the proposed row: TRMT10C does not do the
chemistry, and the ten nuclear protein subunits in the same position all carry the term with
this qualifier. Neither mitochondrial accessory subunit carries GO:0004526 in GOA, so they
are the outliers.

The complex assignment above is a separate query from the qualifier table, and it is the one
this correction turned on, so it is logged rather than asserted:

curl -H 'Accept: application/json' \
  'https://www.ebi.ac.uk/QuickGO/services/annotation/search?geneProductId=UniProtKB:O15091&goId=GO:0030678&limit=20&goUsage=exact'

4 annotations, all PRORP | GO:0030678 | part_of, by IBA, IDA, IPI and TAS. So PRORP is a part
of the mitochondrial ribonuclease P complex, which is what makes it the counterpart rather
than a member of the nuclear comparator set.

The complex column for the 25 protein and ncRNA gene products in the table - the ten nuclear
subunits, PRORP and the fourteen RNAcentral ids, which is every entity in it but the two
ComplexPortal complexes - comes from a third query, run 2026-10-12, because the
qualifier query above returns gene product, qualifier and evidence and says nothing about
complex membership - so filling that column from it would be an inference rather than a
record:

curl -H 'Accept: application/json' \
  'https://www.ebi.ac.uk/QuickGO/services/annotation/search?goId=GO:0005655&taxonId=9606&limit=100&goUsage=exact'

run for GO:0005655, GO:0030677 and GO:0030678, plus a per-id
geneProductId=RNAcentral:URS..._9606 query for each of the fourteen ncRNA ids. Results:
all ten protein subunits are in GO:0005655, four of them also in GO:0030677; PRORP is in
GO:0030678 only; and all fourteen ncRNA ids are in GO:0030677 only, each by part_of IEA.

The ancestry among the three complex terms is from the ontology rather than from QuickGO:

uv run runoak -i sqlite:obo:go relationships GO:0005655 GO:0030678 GO:0030681

GO:0005655 is_a GO:0030681 is_a GO:0030677 and GO:0030678 is_a GO:0030677, which is where the
two-levels-versus-one asymmetry above comes from.

The nuclear subunit genes have no folders in this repository, so these tables are the only
in-branch record of the three queries logged above - goId=GO:0004526,
geneProductId=UniProtKB:O15091&goId=GO:0030678, and the complex-term family - plus the
per-id ncRNA queries; re-run the curls above to check them.

What HSD17B10's review actually says

HSD17B10/MRPP2 is the other non-catalytic subunit of the mitochondrial complex, and its
review is not a stub despite status: INITIALIZED — 101 reviewed rows (49 ACCEPT, 36
KEEP_AS_NON_CORE, 16 MARK_AS_OVER_ANNOTATED, none unreviewed), three authored core functions,
and proposed_new_terms: []. Its third core function models the moonlighting role as
GO:0000049 tRNA binding plus in_complex GO:0030678, which is a recorded judgment to argue
with rather than an absence. Under the comparator above that review warrants the same
contributes_to proposal. TRMT10C also carries GO:0000049, so the two reviews differ only
in this one proposed row.

The organelle-level localization rows

The eight GO:0005739 mitochondrion rows are ACCEPT, and that is a decision rather than an
oversight. TRMT10C (MRPP1) is a mitochondrial matrix protein and GO:0005759 is accepted here on
three rows, so a reader applying the "correct but general terms are MODIFY" rule mechanically, or
re-running an ancestor-based scan, will stop on these. They are not reached by that rule: none of
the eight reasons says the term is less specific than the evidence supports — they give the proteome
and localization evidence and the gene's function. Nothing else in this file refines a
mitochondrial-location ancestor, so the file does not treat one argument two ways. Whether GOA's
organelle-level rows should be refined wherever a compartment row exists is a real question, and it
is a slice-wide one rather than anything specific to this gene.