Taxon Pathway Variance

Why this project exists

A pathway named after its canonical, usually vertebrate, form does not exist in the
same shape in every lineage. Components get lost, replaced, or rewired, while the
families they belong to persist. When that happens, GO terms that presuppose the
canonical wiring keep flowing into lineages where the wiring is absent: by IBA from a
deep PAINT node, by ortholog projection, or by InterPro2GO. Each such annotation is a
small error, but they share a single cause, and a single fix often blocks all of them:
a GO taxon constraint (never_in_taxon / only_in_taxon) on the term, or an IRD
on the lineage in PAINT.

This project does two things:

1. Documents lineage-specific pathway variants — where a pathway runs without a
   component its GO term presupposes, how the lineage does it instead, and which
   modules and gene reviews capture that.
2. Captures proposed new taxon constraints in a machine-readable YAML file
   (proposed_taxon_constraints.yaml),
   each with the evidence for the absence and the annotations it would block, ready to
   hand to GO.

It is the pathway-level complement to IBA_REVIEW section 14
(lineage-inappropriate transfer), which catalogues the individual bad annotations; this
project asks which constraint would have prevented the whole class.

What counts as a proposal

A row in the constraints table needs all of:

• A term that presupposes a component. The GO definition (or its logical
  definition) must require something the lineage lacks; "receptor signaling pathway
  via JAK-STAT" requires a JAK. A process the lineage merely does differently is not
  enough.
• Evidence of absence, not absence of evidence. A primary source stating the
  component is missing from the genome (or a reproducible bioinformatics check in a
  *-bioinformatics/ folder), with a verbatim quote.
• A check that GO does not already have the constraint (OLS term metadata,
  RO:0002161 never_in_taxon / RO:0002160 only_in_taxon, and inherited constraints
  from ancestors).
• The annotations it would block, so the proposal has a measured payoff, and any
  that it would wrongly block (members of the lineage that do have the component).

Taxon ids and GO ids are verified (OLS / QuickGO / NCBITaxon); none are written from
memory.

Case study 1: JAK-STAT without JAK in C. elegans

C. elegans has two STAT paralogs and no JAK.

"There is no conserved homolog of the JAK kinases in C. elegans" — PMID:28874466

Both worm STATs act in JAK-independent pathways, captured in the module
c_elegans_jak_independent_stat_signaling:

• STA-1 represses antiviral genes in the intestine, genetically downstream of the
  ACK-family kinase SID-3 (PMID:28874466).
• STA-2 drives antimicrobial peptide expression in the epidermis after wounding or
  fungal infection, downstream of hemidesmosome (MUP-4) and PMK-1 signalling with
  SNF-12.

The canonical, JAK-based decomposition is jak_stat_signaling,
which now carries a pointer to the worm variant.

JAK-presupposing terms nevertheless reach worm genes by IBA from two families:

The same STAT node also carries GO:0006952 defense response, which reaches STA-1
with the wrong sign: STA-1 represses antiviral genes (IBA_REVIEW section 9).

What the family reviews show

The two family reviews explain why every one of these annotations arrives, and
why a taxon constraint rather than a PAINT edit is the durable fix:

• JAK (PTHR45807) is an animal-only family. PANTHER roots it at Eumetazoa, and
  the tree has no nematode, amoebozoan, fungal or plant leaf. No JAK-STAT term
  reaches a JAK-less organism from this family.
• STAT (PTHR11801) is much older. Its JAK-STAT, defense response and
  proliferation IBDs sit on PTN000927860, the Unikonts node, which spans the
  Dictyostelium STATs as well as all animals. The family review rates the
  JAK-STAT row TOO_DEEP: the seeds support only Bilateria (PTN000210449).
• So the term is placed where the STAT family arose, while the pathway it names
  arose with the JAK family, much later. Any lineage that kept STATs but never had
  JAKs inherits it. That mismatch is structural, which is why it recurs across worms
  and slime moulds and why a constraint on the term catches all of it at once.

The STAT review also found the reverse case at the family root: plant STAT-like
proteins (Arabidopsis SHA/SHB) lack the STAT DNA-binding domain yet inherit the
root's DNA-binding transcription factor IBDs; they are recorded there as member
exceptions.

Case study 2: STAT without JAK in Dictyostelium

Dictyostelium has four STATs (dstA–D) and no JAK; STATc is tyrosine phosphorylated
by the TKL kinase Pyk2 instead.

"There are no orthologs of the animal tyrosine kinases" — PMID:22699506

The statA and statC reviews already MODIFY the inherited GO:0007259 IBA to
GO:0097696 cell surface receptor signaling pathway via STAT, which presupposes no
JAK and is the right term for both the slime-mould and the worm STATs.

Both Dictyostelium STAT arms (STATc downstream of DIF-1 and hyperosmotic stress via
the TKL kinases Pyk2/Pyk3 and the PTP3 phosphatase gate; STATa downstream of cAR1 in
the prestalk tip) are modelled in
dicty_jak_independent_stat_signaling.

The PIAS case is also where the repository gained member exceptions in family
reviews: PANTHER subfamily PTHR10782:SF94 holds both the fly PIAS (a genuine JAK-STAT
regulator) and worm GEI-17, so no subfamily-level scope can separate them. See
interpro/panther/PTHR10782/PTHR10782-review.yaml.

Proposed taxon constraints

The case for TPV-001 to TPV-003, with node-level PAINT recommendations, is written up
for GO editors and PAINT curators in the public briefing
JAK-STAT Without JAK.

See proposed_taxon_constraints.yaml.
Each proposal records the GO term, the constraint (never_in_taxon / only_in_taxon),
the taxon, any constraint GO already has, verbatim evidence, the annotations it would
block, any it would wrongly block, a status (PROPOSED / SUBMITTED / ACCEPTED /
REJECTED), and where in this repo the case was found.

Current proposals

TPV-004 to TPV-007 tighten constraints GO already has (Eukaryota, never in Fungi, Bilateria); each
YAML entry records the existing one. TPV-008 adds a constraint to a term that has none. They come
from the RAR_MISANNOTATION re-examination of Wei et al. 2020, and the counts
are exact-term QuickGO annotations on 2026-11-05 (TPV-008: 2026-11-07).

Outcome for TPV-001 to TPV-003

All three were requested in
go-ontology#32689
(2026-10-05), with the briefing as the supporting analysis. GO added them on
2026-11-03: neverin:6231 on GO:0007259 and GO:0046425 in commit
98389d1
(PR #32857),
and onlyin:33208 on GO:0007259 in commit
fed722d
(PR #32861) after a reviewer
noticed it had been missed. The Nematoda constraints were taken at phylum level, not the
Caenorhabditis fallback. Commit 98389d1 also renamed GO:0046425 and its positive and negative regulation children
to insert "cell surface" (go-ontology#14167). Each proposal's submission block in the YAML records the issue,
the commits and the dates.

One loose end: GO applied the Metazoa constraint without the TPV-003 prerequisite
(re-parenting GO:0007260), so GO:0007260 now inherits it by part_of. GO:0007260 is
already listed for obsoletion in
go-ontology#27064, which
would remove the conflict. If it stays, request the re-parenting there.

To do

• [ ] Follow go-ontology#27064.
  If GO:0007260 is not obsoleted, ask there for it to be re-parented under GO:0097696
  (the TPV-003 prerequisite), since it now inherits only-in Metazoa by part_of.
• [ ] Gene reviews for the six members of
  dicty_jak_independent_stat_signaling
  that have none yet: Pyk2 (splB, UniProtKB:P18161), Pyk3 (UniProtKB:Q54I36),
  PTP3 (ptpC, UniProtKB:P54637), Phg2 (UniProtKB:Q54QQ1), DrkA (UniProtKB:Q54H46)
  and GskA (UniProtKB:P51136). These complete the module's gene-review coverage; Pyk3
  needs particular care because its effect on STATc is disputed between strains
  (PMID:18657170 vs PMID:25143406).
• [x] sta-2 OpenScientist result wired in: STAT DNA-binding-domain fold kept, sequence-reading
  features not; GO:0000978 marked over-annotated, DNA-binding TF terms generalised to
  GO:0140110 transcription regulator activity.
• [x] gei-17 OpenScientist result wired in: it rules out a SAP-domain binding route, but
  GO:0140416 also covers inhibition by SUMOylation, which GEI-17 may do to the piRNA USTC
  complex (PMID:40316696, abstract only), and GEI-17 keeps PIAS1's SIM-adjacent acidic module,
  so a ligase-independent binding route is not excluded either; GO:0140416 is now UNDECIDED.

Candidate cases to scope next

• TYK2 IRD. PAINT records TYK2 as having lost JAK-STAT signaling (a negated IRD
  on PTN002910252); the JAK family review rates this WRONG_NODE, contradicted by
  TYK2's own experimental annotations. Not a taxon question, but the same
  machinery misfiring in the opposite direction.
• Other pathways named for a component with a patchy distribution, collected as they
  surface in IBA and family reviews.

Key literature

Reviews that frame the JAK-STAT case:

• Liongue et al. 2013, Evolution of the JAK-STAT pathway (PMID:24058787). Where
  JAKs and STATs arose; the main source for TPV-003.
• Wang & Levy 2012, Comparative evolutionary genomics of the STAT family of
  transcription factors (PMID:24058748). STATs across eukaryotes, including amoebae
  and nematodes.
• Kawata 2011, STAT signaling in Dictyostelium development (PMID:21534947). STAT
  activation in a lineage without JAK.
• Taffoni & Pujol 2015, Mechanisms of innate immunity in C. elegans epidermis
  (PMID:26716073). Context for the STA-2 pathway.

Primary evidence of JAK absence: Tanguy et al. 2017 (PMID:28874466) for C. elegans;
Goldberg et al. 2006 (PMID:16596165) and Araki et al. 2012 (PMID:22699506) for
Dictyostelium.

Related

• IBA_REVIEW — section 14, lineage-inappropriate transfer
• GENOME_WIDE_VALIDATION — reuses GO taxon constraints as
  a consistency check
• PAINT — family reviews and node assessments