Design Strategy (Physics-Based, No AI)
- We used an in house deterministic Global Minimum Energy Conformation (GMEC) solver for fixed-backbone protein design.
- Rosetta provided one- and two-body rotamer energies (Dunbrack library); our optimizer used these matrices to explore low-energy sequences far faster than Rosetta’s packer.
Structural Rationale
- Backbone taken from chain B of 2VSM, representing the ephrin-B2 interface targeted by Nipah G.
- EFNB2 is large and membrane-bound; this fragment likely needs stabilization, hence we redesigned its full sequence, i.e., we allowed all amino acids at all positions with full Dunbrack rotamers.
Complex-Based Evaluation
- Energies were computed for the relaxed (FastRelax) 2VSM complex.
- Only the receptor chain was allowed to mutate; the virus protein chain remained fixed.
Iterative Design & Ranking
- First round: hundreds of sequences ranked by GMEC energy and ipSAE.
- Top candidates were fed into Boltz2, then redesigned again based on the predicted complex, yielding new sequence sets.
- Final selections ranked by ipSAE.
- An N-terminal methionine was added to improve expression.
Submission Note
- Due to inconsistencies between local and platform ipSAE, we submitted one top sequence to maximize expression chance.
- In retrospect, a larger set may have been better.
In summary, this is a physics-driven, GMEC-optimized redesign of the EFNB2 binding interface, chosen for biological relevance to Nipah G.
