Method description for RBX1 binder design

1. Background RBX1 (Ring-Box Protein 1, also known as ROC1) is the core catalytic component of the Cullin-RING E3 ubiquitin ligase (CRL) complex. Through its C-terminal RING domain, it recruits E2 ubiquitin ligases and plays a central role in key biological processes such as cell cycle regulation and DNA repair. We selected the conserved E2 ubiquitin ligase-binding surface/native inhibitory interface on the RING domain of RBX1 as the focus of our study.
2. Methodology 2.1 Structural Analysis and Target Selection We analyzed the following RBX1 complex structures to identify optimal design targets. PDB 4F52 (Glomulin-RBX1-CUL1): Reveals that the endogenous inhibitory protein Glomulin exerts its inhibitory effect by tightly binding to and completely masking the E2-binding surface of RBX1. This interface (880 Ų) contains distinct hydrophobic clusters and key polar-polar interactions. PDB 4P5O (RBX1-CUL1-E2NEDD8): This structure displays the conformation of RBX1 in its active state. It shows the key binding sites between RBX1 and the E2 ubiquitin ligase where the unique Arg46 “central peg” plays a central role. Identification of Hotspots: Based on an interaction analysis of the RBX1 complex, we selected eight residues on the RBX1 RING domain that are critical for binding affinity and specificity as design anchors: A35, I36, R38, E47, Q49, W79, P87, and L88. These residues constitute the core hydrophobic cluster and the key electrostatic interaction network at the binding interface. 2.2 Generation of initial binder candidates In this step, we employed multiple independent generation methods, including Boltzgen, Protenix, BindCraft, and PPIFlow, to generate de novo binders of RBX1. All four methods use the same RBX1 structure (PDB:2LGV) and the hotspots identified in 2.1 as input. 2.3 Assessment of Affinity and Binding Quality We used AlphaFold2 Multimer to predict the complex structures for each “design protein–RBX1” pair. Preliminary filtering: Remove any sequences predicted to be unstable by the model itself (e.g., those with a pLDDT < 80 in AlphaFold2 monomer predictions) or those with clearly unreasonable complex structures. Core ranking metric: We performed a global descending rank based on the predicted interface ipTM scores. Supplementary validation: We used chimeraX to verify that the designed protein forms effective spatial interactions with the majority of the predefined hotspot residues in the predicted complex, thereby ensuring that the design targets the correct functional epitope.
3. Final list of candidate designs Through the aforementioned parallel generation and unified screening process, we ultimately selected the 88 candidate ligands with the best computational predictions from thousands of initial designs. The top 10 designs exhibited excellent computational metrics, with a predicted ipTM > 0.88—significantly higher than that of the other candidates. In addition, we investigated and confirmed that the positively charged residues on the designed protein can form a stable salt bridge with Glu47 of RBX1 and interact with the key site R38.