Worked example: RPTOR / Raptor (Q8N122)

Back to the project page.

Third worked example and the second driving a new review of a previously
unreviewed gene (after LDLRAP1). RPTOR is the
prototype of the GO:0140767 enzyme-substrate adaptor activity child — the
case the project's modelling rule singles out, where the adaptor's job is to
bring an enzyme together with its substrate (provides input for), not to
regulate a partner's activity. See the
full review of RPTOR (linked from the gene symbol) and its notes at
genes/human/RPTOR/RPTOR-notes.md.

The gene in one line

RPTOR (Raptor) is the non-catalytic substrate-presenting subunit of mTORC1.
UniProt states it plainly: within the complex RPTOR acts both as a recruitment
adapter (binding the Rag GTPases to dock mTORC1 at the lysosome) and as a
substrate-specific adapter that binds TOS-motif proteins (4E-BP1, S6K1)
and directs them towards the active site of the MTOR kinase domain
(UniProtKB:Q8N122, CC:FUNCTION; mechanism in PMID:12747827, PMID:24403073).

Starting state (GOA, 170 rows)

• 61 × GO:0005515 protein binding (IPI) — the largest block, almost all
  from interactome/affinity screens.
• The real adaptor MFs present: GO:0140767 enzyme-substrate adaptor activity
  (IDA ×3), GO:0030674 protein-macromolecule adaptor activity (IDA/IBA),
  GO:0031267 small GTPase binding (Rag), GO:0019901 protein kinase binding.
• A large tail of mTORC1 downstream processes (growth, autophagy, glycolysis,
  lipid synthesis, Pol III transcription, nutrient responses) and some
  tissue-/behaviour-level ISS rows.

What the decision rules produced

170 annotations → ACCEPT 82, MODIFY 30, REMOVE 32, KEEP_AS_NON_CORE 22,
MARK_AS_OVER_ANNOTATED 4.

The instructive part is how the 61 protein binding rows split by partner
identity — the single most useful move on an adaptor. The key discipline: a
binary interaction (an IPI hit) evidences binding, not the bridging
activity, so each row is refined to a binding term the partner supports
(or removed), and the GO:0140767 enzyme-substrate adaptor activity is left to
the three IDA rows that actually assay the bridging:

Interaction partner	Rows	Call	Term
MTOR kinase (P42345, + mouse ortholog)	24	MODIFY	GO:0019901 protein kinase binding (the enzyme side)
S6K1 / RPS6KB1 (P23443), a protein kinase	3	MODIFY	GO:0019901 protein kinase binding
4E-BP1 (Q13541), PRAS40 (Q96B36) — TOS-motif substrates, not kinases	8	REMOVE	bare protein binding (no specific binding term; the adaptor activity is already asserted by IDA)
everything else (MLST8, BRAF, singletons)	24	REMOVE	bare protein binding, per policy

(The rows sum to 59, not 61: the 61 GOA GO:0005515 lines collapse to 59
distinct reference+partner entries in the review, as with GRB2.)

That triage reconstructs the mechanism without over-claiming: RPTOR binds the
MTOR kinase and the kinase substrate S6K1 (→ protein kinase binding), and
its binding of the non-enzyme substrates 4E-BP1/PRAS40 is recorded as the
enzyme-substrate adaptor activity only where an assay (IDA) demonstrates the
bridging — not inferred from a two-hybrid hit.

Core functions settled on the two adapter roles: GO:0140767 enzyme-substrate
adaptor activity (substrate presentation to MTOR) and GO:0030674
protein-macromolecule adaptor activity (Rag-GTPase-mediated recruitment to the
lysosomal membrane), both in_complex GO:0031931 TORC1 complex.

What this example adds beyond GRB2 and LDLRAP1

• It exercises the enzyme-substrate / provides input for pattern. GRB2 and
  LDLRAP1 are regulatory/recruitment adaptors; RPTOR is the substrate-presentation
  case the modelling guideline treats separately. In a GO-CAM, RPTOR's
  enzyme-substrate adaptor activity has input the TOS-motif substrate and
  provides input for the MTOR kinase activity — it does not regulate it.
• Partner identity does the work — but only for the binding term. Sorting a
  61-row protein binding pile by interactor into "the enzyme" vs "the
  substrates" vs "noise" is faster and more defensible than reading 61 papers. It
  yields the right binding refinement (protein kinase binding) or a REMOVE; it
  does not license asserting the bridging activity (GO:0140767) from a
  binary hit — that stays with the IDA rows that assay it. An earlier draft
  MODIFYed eleven substrate-binding rows to the activity term; that over-claimed
  and was corrected to three protein kinase binding refinements plus eight
  removals.
• Over-annotation on a hub gene is mostly pleiotropy, handled by
  KEEP_AS_NON_CORE, not REMOVE. mTORC1 touches growth, autophagy and metabolism;
  those experimental process rows are true and were kept non-core. Only distal
  ISS/IEA rows (social behaviour, osteoclast/odontoblast differentiation,
  xenobiotic response) were marked over-annotated — consistent with not
  overruling experimental curators.
• contributes_to is the right qualifier for a non-catalytic subunit. RPTOR's
  contributes_to GO:0004674 protein serine/threonine kinase activity was kept:
  it credits the subunit for the complex's catalytic activity without claiming
  RPTOR is itself a kinase.

Caveat

RPTOR is a heavily-curated hub (170 annotations, 91 references, most not in the
local publication cache), so many experimental ACCEPT rows are deferred to the
original curators without an added verbatim quote (the validator emits advisory
supported_by warnings for these). The core-function claims and the partner-based
MODIFY calls are grounded in cached, quote-verified references.