Conidiation regulatory cascade — module design proposal

Status: reusable ABSTRACT module implemented. modules/conidiation_regulatory_cascade.yaml
(rendered: pages/modules/conidiation_regulatory_cascade.html) is now an ABSTRACT
module with an EXACTLY_ONE taxon variant_set holding two paradigms:

• aspergillus_paradigm — the fully-curated FluG/Flb → BrlA → AbaA → WetA/velvet
  cascade with G-protein/FlbA gating; all 21 member genes have validated reviews
  under genes/EMENI/. Carries the conidiophore-development (GO:0070787) structural concept.
• neurospora_macroconidiation — the analogous, largely non-orthologous N. crassa
  program, grounded in verified UniProtKB exemplars and now fully reviewed (genes/NEUCR/):
  the blue-light White Collar Complex (WC-1 Q01371, WC-2 P78714) + circadian FRQ (P19970)
  gating, the fluffy Zn(II)₂Cys₆ master TF FL (O13360), and structural output (EAS/CCG-2
  hydrophobin Q04571, CON-6 P34762, CON-10 P10713).

Root concept generalized to conidium formation (GO:0048315); context broadened to
Pezizomycotina. This page remains the design rationale and the source for the two-axis
ontology analysis and the annotation-inconsistency caveat.

Implemented so far: the six-tier cascade (FluG/Flb → BrlA → AbaA →
WetA/velvet → structural output, plus G-protein/FlbA repressive gating), dual
top-level concepts GO:0048315 + GO:0070787, per-annoton molecular functions, and
typed connections.

Member gene reviews — 21/21 complete (genes/EMENI/); every module-grounded
gene now has a validated review with core functions:

Module MF verification (from reviews): the velvet structural paper
[PMID:24391470] confirms VosA/VelB carry an NF-κB-like DNA-binding domain, so the
module's GO:0003700 on those annotons holds; WetA (not a velvet protein) had its
MF removed; LaeA's GO:0008168 methyltransferase is confirmed by IDA.

Status: fully reviewed and PTN-grounded. Both variants are complete — 28/28 grounded
genes have validated reviews (21 EMENI + 7 NEUCR; con-10 has no GO annotations to complete),
module deep research is present and its citations are now checkable
(conidiation_regulatory_cascade-deep-research-manual.md; the four previously-uncached PMIDs it
relies on — PMID:9529886, PMID:7750148, PMID:8617205, PMID:16387865 — are cached in
publications/), and 18 conserved-role annotons now
carry PANTHER PTN ancestral-node grounding (FAMILY selectors with a PTHR family term, a verified
UniProtKB representative, and PAINT PTN… nodes resolved from the local IBD index + GOA WITH/FROM,
GO_REF:0000033). Nodes propagating review-flagged terms were excluded (flbD cell-cycle node,
wA oxidoreductase node); the brlA node/descriptor family split (PTN in PTHR16515 vs family
PTHR14003) is recorded.

The Neurospora stage genes acon-2 and acon-3 were investigated and added
(module now 30/30 reviewed): acon-2 (U9W570/NCU00478) is a cAMP
phosphodiesterase (confirmed by PMID:31172300) acting upstream of FL — added to the
gating tier with PTN grounding (PTHR11347, PTN001682918); acon-3 (Q7SB37/NCU07617,
"Acr1") is a nuclear MEDUSA/MedA-family (PTHR39463) regulator downstream of FL — a new
stage-regulation tier. The epistasis acon-2 → FL → acon-3 was confirmed verbatim from the
primary source (PMID:15126394); that source does not extend the chain to the con genes —
it reports fl-driven morphogenesis correlating with eas induction but not con-6 or
con-10, so the downstream acon-3 → con step remains uncited and is flagged as such in the
module.

1. What the module is

Conidiation (conidiogenesis) is the developmental program that produces
conidia — asexual, mitotically-derived spores borne on specialized aerial
structures (conidiophores) in filamentous ascomycetes. It is one of the
best-dissected fungal developmental programs, worked out chiefly in Aspergillus
nidulans and Neurospora crassa.

The reusable, defensible module is the central regulatory cascade — the
transcription-factor relay that commits vegetative hyphae to sporulation and
drives spore maturation — together with the signaling that gates it and the
structural genes it ultimately switches on.

2. Module boundary

Type: module_type: DEVELOPMENTAL_PROCESS.

The four confirmed inclusions beyond the BrlA→AbaA→WetA spine — upstream
activation (FluG/Flb), repressive gating (FlbA / G-protein–PKA), velvet
maturation (VosA/VelB), and structural output (hydrophobins / pigment) —
are all modeled as first-class tiers rather than prose context.

3. Top-level grounding

Resolved. Give module.concepts two terms (see §3a for why): the
spore-cell axis GO:0048315 "conidium formation" (primary) and the structural
axis GO:0070787 "conidiophore development". The cascade drives both.

⚠️ Do not use GO:0061794 "conidium development". It is being obsoleted
as an unnecessary grouping term (0 direct annotations, single child
GO:0048315) — GO tracker
geneontology/go-ontology#32315
(opened 2026-07-15, label obsoletion; still isObsolete:false in the
released ontology as of this writing, i.e. an in-flight change). GO:0048315
is the surviving specific term.

Related terms to use where appropriate rather than at the top:

• GO:0048315 conidium formation — top-level module concept (the whole
  program, formation → mature spore).
• GO:0075306 regulation of conidium formation — for the regulatory-cascade
  framing / the FluG–Flb–BrlA–AbaA–WetA relay as regulators.
• GO:0030436 asexual sporulation — broader parent (context only).

Per-tier MF/BP terms (TF activity, RGS activity, PKS activity, etc.) still go on
the leaf annotons and are resolved via OLS during grounding.
- module.context:
- taxa: Pezizomycotina / Ascomycota (species-neutral at the top; species
pinned inside variant_sets). Resolve NCBITaxon id.
- cellular_components: nucleus (the TF relay), plasma membrane / hyphal tip
(signal sensing), extracellular region (FluG signal, rodlet layer). Avoid
asserting both a parent and child compartment without a recorded reason.

3a. Ontology landscape (scouted) — two orthogonal axes

Conidiation splits into two GO branches that are not parent/child. Both
belong in the module — spore-cell formation as the primary concept, conidiophore
development as a co-equal structural concept/context.

* Indicative QuickGO goUsage=exact counts — not a basis for term choice
(see caveat below).

⚠️ Annotation is inconsistent — do not infer structure from GOA. 9283
annotations sit on the vague grouping term asexual sporulation while the
precise conidium formation (551) and conidiophore development (19 / 0 / 1)
terms are sparsely populated. The same regulators (brlA, abaA, wetA, …)
are annotated to a scatter of broad and specific terms depending on the
MOD/curator. Consequences for the build:

• Choose each annoton's term by biology, not by copying where the gene
  currently sits in GOA.
• Expect many existing broad-term annotations (e.g. to GO:0030436) to be
  MODIFY candidates toward the specific conidium/conidiophore terms when the
  member gene reviews are done.
• Use regulation sub-terms (GO:0075306 / GO:0070793 and their pos/neg children)
  for the TF-cascade tiers where the biology is regulatory, and the formation/
  development terms where it is executive.

4. Part decomposition

Ordered parts, each a ModuleNode (REGULATORY_STEP, except the output tier)
holding leaf annotons for the member proteins.

Molecular-function terms (TF activity, RGS/GTPase-regulator activity, PKS
activity, structural constituent) go on the leaf annotons, never on the
process module concept — the standard process-module modeling rule.

5. Cascade diagram

6. Connections (typed edges)

Using ModuleConnectionTypeEnum:

Regulatory/developmental edges take chaining_status: NOT_APPLICABLE if the
advisory reaction-continuity check ever flags them (this is not a metabolic
chain).

7. Species variation — variant_sets

The Aspergillus and Neurospora programs are alternative implementations of
the same developmental logic. Model tiers 1–4 with a variant_set on the
taxon/lineage axis (EXACTLY_ONE) so the module stays reusable:

• Variant A — Aspergillus paradigm: BrlA → AbaA → WetA, with FluG/FlbA–E
  upstream and velvet (VosA/VelB/VeA/LaeA) maturation.
• Variant B — Neurospora macroconidiation: FL (Gal4-type Zn₂Cys₆ TF),
  ACON-2/ACON-3, with White-Collar-Complex (WC-1/WC-2) + frq circadian
  gating; EAS hydrophobin and con-6/con-10 as structural output.

Ground each variant with its own representative_members; do not inflate the
member list to every species named in deep research.

8. Member roster & grounding TODO

Resolve each of these during the build (UniProt accession; PANTHER family/PTN
where the local cache has one; per-annoton GO MF/BP term). Reference organism
dirs: likely EMENI (A. nidulans) and NEUCR (N. crassa).

9. Anti-patterns to avoid (from the module-curation skill)

• Putting a member's MF term on the process module concept instead of its leaf
  annoton.
• Letting the module collapse to a species-specific A. nidulans member list —
  keep it reusable via variant_sets + representative_members.
• Asserting parent+child compartments (e.g. cytoplasm and cytosol) without a
  recorded reason.
• Treating deep-research prose as an identifier source — every id resolved
  against UniProt / PANTHER / OLS / GO-CAM.

10. Build workflow (when approved)

1. just fetch-gene EMENI brlA (and each core member) → seeds gene reviews +
   UniProt/GOA; repeat for NEUCR variant members.
2. just module-deep-research-perplexity conidiation_regulatory_cascade →
   cited modules/conidiation_regulatory_cascade-deep-research-*.md.
3. Author modules/conidiation_regulatory_cascade.yaml per the skeleton above.
4. Validate + render:
5. uv run linkml-validate -s src/ai_gene_review/schema/gene_review.yaml -C ModuleReview modules/conidiation_regulatory_cascade.yaml
6. uv run python -m ai_gene_review.validation.module_validator modules/conidiation_regulatory_cascade.yaml
7. just render-module modules/conidiation_regulatory_cascade.yaml
8. If this project page is kept, add a project-card entry to
   pages/projects/index.html (the index is manually maintained).

11. Open questions

• Single reusable module vs. two concrete instances? Recommendation:
  one reusable module with taxon variant_sets (Aspergillus + Neurospora).
• ~~Which top GO term best spans the whole cascade?~~ Resolved:
  GO:0048315 conidium formation (see §3; GO:0061794 is being obsoleted).
• Structural output — in-core vs. sibling module? Currently in-core as
  tier 5; could be spun out to a conidial_wall_assembly module if it grows.
• GO-CAM coverage: check gocams/index.tsv for any existing conidiation
  models to attach via gocam_associations.