The high-affinity protein binders against the Nipah virus attachment glycoprotein (NiV-G) were designed using BoltzGen, a diffusion-based de novo protein design model. The design process was guided by a set of key epitope residues S241, R402, E501, W504, E505, V507, Q530, E533, and N557, identified through published literature and molecular dynamics (MD) simulations of NiV-G in complex with human cell surface Ephrin B2 receptor (PDB: 2VSM). These analyses revealed residues involved in persistent or energetically significant interactions, making them ideal targets for rational binder engagement.

BoltzGen was configured to generate nanobody-like binders and larger synthetic protein scaffolds (<250 amino acids), to explore diverse structural topologies. The model was conditioned on the selected epitope to optimise shape complementarity, chemical complementarity, and interface packing.

Designed binders were further validated using Boltz2 (https://atharvatilewale.github.io/boltz2-notebook/) for binder folding and binder–target complex modelling. Candidates were initially filtered based on the RMSD between the designed and predicted structures. Models exhibiting low RMSD values were further evaluated using the ipSAE interface metric to prioritise binders with strong predicted structural and interaction quality.

The overall workflow of design is as follows:

MD Simulation → Interacting Residue Identification → De Novo Binder Design → Structural Prediction (Boltz2) → RMSD Filtering → ipSAE Scoring → Final Binder Selection