Methodology

The target structure was derived from PDB ID [e.g., 3DPL, 1LDJ]. Design constraints were defined based on structural analysis and prior studies of Cullin-RING ligase (CRL) assembly. The RING domain was constrained to preserve native zinc coordination and key hydrophobic interactions. Given the conformational flexibility of the RING finger, a critical constraint was to enable stabilization through an intermolecular β-sheet interaction at the Cullin C-terminus, while avoiding steric clashes with the zinc-binding site.

Helical binder for RING domain - BindCraft

We utilized BindCraft, a structure-guided protein binder design framework, to generate sequences targeting the RING domain interface. Secondary Structure Bias: We applied structural priors to favor helical binders complementary to the RING domain interface, thereby stabilizing the flexible RING finger while ensuring no perturbation of its native zinc-coordination. Distance-Based Hot Spot Selection: Hot spots were defined based on residue-level distance analysis from reference RBX1 and CRL complex structures, identifying interface regions most suitable for productive binder engagement. Literature-Guided Hot Spot Validation: To preserve native RING-domain binding characteristics, the selected hot spots were cross-validated against reported RING binders and known CRL assembly interfaces, ensuring compatibility with established structural interaction modes. Size and Diversity: Binder lengths were set at 80-110 residues to provide sufficient surface area for binding while maintaining a compact, stable fold.

Disordered beta binder BoltzGen

De novo protein binders targeting the intrinsically disordered N-terminal region of RBX1 (residues 1-18; PDB: 3DPL) were designed using BoltzGen, a diffusion-based generative framework. Binders were constrained to 70-100 residues with no sequence pre-specification. From 3,000 generated candidates, 747 designs (24.9%) passed structural self-consistency and sequence quality filters, with backbone RMSD below 2.5 Å in both standalone and complex refolding (mean 2.05 and 1.82 Å, respectively), mean design-to-target ipTM of 0.335, and average buried interface area of 1,743 Ų. Secondary structure composition was diverse (mean: 35% helix, 27% sheet, 38% loop), with 119 designs (15.9%) exhibiting sheet-dominant character consistent with the intended β-binder topology.

Bispecific protein fusion

Sequences were designed using BindCraft and BoltzGen. To generate a bispecific protein fusion from two designed domain binders, a linker-based fusion strategy was applied. To ensure independent folding and functionality of each domain, we:

Introduced a protein linker to connect the two domains while minimizing steric interference. Optimized linker length and composition to maintain solubility and reduce aggregation. Employed primarily flexible linkers (e.g., Gly-Ser rich) to provide conformational freedom between domains.

In Silico Validation and Filtering Protocol The designed candidates underwent a rigorous multi-step evaluation:

ΔG Calculation: Estimated binding free energies for both trimmed domain binders and full complex structures using an MM/GBSA-based approach. Zinc ions (Zn²⁺) Filtering: Designs that caused significant RMSD shifts in the RBX1 RING domain (especially the zinc-binding loops) were discarded to ensure the binder does not effect the zinc coordination. Topological Filtering: A Topology filter was applied to ensure physically realizable binder-target complexes at β-sheet regions.

Final Selection for Experimental Testing A final library was selected for experimental validation. This library represents a balance between diverse topologies and high-confidence $\beta$-sheet stabilizers and RING domain stabilizers, aiming to provide a robust solution for the challenging RBX1 target. Notably, our design framework explicitly modeled the three-zinc-coordinated state of the RING domain to reflect realistic physiological conditions; therefore, it is highly recommended that subsequent experimental assays maintain this specific metal stoichiometry to accurately mirror the structural constraints established during the in silico design process.