De novo miniprotein can block the IL-17RA-contact face of human IL-17A (4HSA)

Claim: A de novo miniprotein designed on the BIOS production stack against the IL-17RA-contact face of human IL-17A (locked PDB 4HSA chain A; mature epitope Y85, H86, Q94, E95, L116 and C-terminal I127–V131) can form a soluble, epitope-competitive binder that reduces IL-17A–driven IL-6 induction in a responsive cell assay in a dose-dependent manner, preferentially versus IL-17F.
Reasoning: The IL-17A/IL-17RA crystal structure (PDB 4HSA) defines a receptor-contact patch suitable for steric blockade; therapeutic antibodies validate that blocking this face is biologically meaningful. Miniprotein generators (PXDesign, BoltzGen, RFdiffusion3) plus co-fold/novelty gates are intended to produce compact binders to that patch without Cys staples unless requested.
Test / falsification:
- Express/purify the top shortlisted design; confirm SEC monomer + intact mass.
- SPR/BLI: saturable binding to IL-17A that is competed by IL-17RA-Fc or an RA-blocking reference antibody. Failure = no saturable binding or no competition (wrong face).
- Functional: dose-dependent reduction of IL-17A→IL-6 (or equivalent). Failure = no functional block at highest tested concentration.
- Selectivity kill: IL-17F binding ≥ IL-17A under the same assay → claim fails.
Evidence (structure / rationale only — not wet-lab proof of this design):
- Liu et al., Nat Commun 2013: IL-17A / IL-17RA complex, PDB 4HSA (https://doi.org/10.1038/ncomms2880 ; https://www.rcsb.org/structure/4HSA).
- BIOS staging run d5a88899 (chat b21781e2-fd15-4583-827a-df363f6b6413): requested 4HSA; panel reported Ran against 4HR9 · chain A → stamped protocol_deviation. BoltzGen returned 0 designs after shard failure with a green check → empty-engine protocol_deviation. No ipSAE/Kd/hit-rate claimed here; ranking ≠ binding.
Uncertainty: Until a 4HSA-locked Full-depth BIOS shortlist is exported with real ensemble/novelty fields (or explicitly UNKNOWN) and wet-lab criteria are run, this remains a design hypothesis, not a demonstrated binder.