To design binders against RBX1, we followed a dynamics-driven and multi-strategy computational approach.

RBX1 is a small RING-domain protein that shows significant flexibility, especially at its terminal regions. Because of this, static PDB structures may not fully represent its biologically relevant conformations. Our hypothesis was that using a dynamically stable conformation of RBX1 would improve binder design quality and realism.

To address this, we first performed a 100 ns molecular dynamics (MD) simulation of RBX1 in an explicit solvent environment to better approximate physiological conditions. The trajectory was analysed using RMSD, RMSF, and Free Energy Landscape (FEL) methods. From this, we selected low-energy, structurally stable conformations to use as targets for binder design.

Binder generation was then carried out using two complementary approaches:

1. BoltzGen (de novo design) Used to generate diverse binder backbones and sequences conditioned on the selected RBX1 conformations, allowing exploration of a broad structural space.

2. RFdiffusion + ProteinMPNN (structure-guided design) RFdiffusion was used to generate backbone scaffolds targeting specific interface regions on RBX1 (based on hotspot residues), followed by ProteinMPNN to design compatible amino acid sequences. Multiple parameter settings (e.g., binder length, hotspot combinations, iterations) were explored to increase diversity and improve binding potential.

All generated designs were filtered using multiple criteria to ensure structural quality and binding relevance. These included: • pLDDT and iPAE/interface confidence • RMSD and structural consistency • Interface quality and contact formation • Hydrophobicity and solubility balance • Removal of steric clashes and unstable folds

Only designs that passed all filtering steps consistently were selected for submission. Overall, this approach combines protein dynamics, generative design, and multi-parameter optimisation to improve the chances of identifying binders that are stable, realistic, and compatible with RBX1’s flexible nature.