STRC h09 Phase 4o AF-M Nanobody Method Transfer 2026-04-27

Harvey 2026 imports as a cautionary secondary interface-score gate for h09; no new VHH therapy and no S-tier movement.

Problem

The Bio+AI Daily post highlighted Harvey et al. 2026: AF-M screening of synthetic VHH/nanobody sequences against GPCR MRGPRX2 found experimentally validated nanomolar binders. The question for STRC is whether this creates a new therapeutic hypothesis or strengthens an existing one.

Method

  • Read the post first, before paper/code.
  • Retrieved and MinerU-parsed the Nature Communications paper and selected supplementary files.
  • Reviewed h09 and h26 hubs/logs because these are the active protein-interface hypotheses.
  • Reviewed the GitHub scorer as a method reference and filed a tool card.

Results

Candidate useDecisionRationale
New STRC nanobody therapyReject / do not registerDelivery, immunogenicity, expression, and target-class risks are high; paper does not validate non-GPCR targets.
h26 homodimer rescueNot actionable nowh26 currently lacks a passing structural candidate; Harvey’s method does not rescue failed non-GPCR homodimer designs.
h09 method importKeep as secondary gateh09 already uses AF3 interface scores and negative controls; Harvey gives a concrete parser/metric recipe and a strong warning not to interpret scores as Kd.

Numbers imported

FindingValueSource
GPCR top-5% precision0.93-1.02026-harvey-afm-nanobody-gpcr
Soluble/non-GPCR top-5% precision0.222026-harvey-afm-nanobody-gpcr
Prospective screen scale10000 VHH sequences2026-harvey-afm-nanobody-gpcr
Above-threshold hits179 / 100002026-harvey-afm-nanobody-gpcr
Validated bindersranks 1, 5, 72026-harvey-afm-nanobody-gpcr
Validated Kd range20-200 nM2026-harvey-afm-nanobody-gpcr
Interface contact cutoff10 A C-alpha distance2026-harvey-afm-nanobody-gpcr

Phase skeleton

Only run this if Phase 4m/4k results are ambiguous enough that a method cross-check is worth the extra work.

PhasePurposeOutputGate
4oLiterature/method importThis noteNo tier movement
4pParser adaptationtail91 AF3/AF-M interface PAE, pLDDT, pDockQ, model support vs matched negative controlsMethod confidence only; cannot promote tier
4qOptional VHH reagent screenIf ever needed, VHHs as assay reagents against TMEM145/STRC, not as therapyRequires wet-lab validation before any hypothesis promotion

Verdict

  • Outcome: METHOD-IMPORT.
  • Interpretation: useful as a discipline layer for interface-score parsing and negative controls; not useful enough to create a new hypothesis.
  • Limitation: h09 is not a GPCR nanobody problem. The paper’s own non-GPCR benchmark is the load-bearing warning.

Ranking delta

  • Hypothesis h09: tier tentative-S tentative-S | mech 5 5 | deliv 4 4 | misha_fit 3 3.
  • Next step unchanged: Phase 4m / Phase 4l / Phase 4k remain the active gate closers.
  • Evidence added: method note and tool card. No tier movement.

Connections