Methodology We designed protein binders targeting an exposed edge β-strand within the RBX1 structure. The selected target region spans residues F103–G107, corresponding to a β-sheet segment that is accessible for intermolecular β-strand pairing, making it a suitable interface for binder design.

Backbone Generation Binder backbones were generated using the RFdiffusion3 framework within the Foundry pipeline. A total of 500 backbone conformations were sampled, conditioned on the geometry of the target β-strand to promote complementary β-sheet extension.

Sequence Design For each generated backbone, four amino acid sequences were designed using ProteinMPNN. This step optimized residue identities to stabilize the binder fold and enhance interactions with the target interface.

Hotspot Incorporation To improve binding affinity, hotspot residues were incorporated based on relative solvent accessible surface area analysis. Residues with favorable solvent exposure and potential for interaction were prioritized to anchor the binder onto the RBX1 β-strand.

Structure Prediction The designed sequences were structurally evaluated using RosettaFold3. This step provided predicted complex structures and confidence metrics for downstream filtering.

Filtering Criteria The generated designs were filtered based on a combination of structural confidence and energetic metrics:

RF3-based metrics: Binder pTM ≥ 0.8 Interface predicted alignment error (ipAE) ≤ 1.5 Rosetta-based metrics: Binding free energy (ΔΔG) ≤ 30 Surface aggregation propensity (SAP score) ≤ 35 Contact surface area ≥ 300 Ų These thresholds ensured selection of binders with high structural confidence, favorable energetics, and sufficient interface packing.