Upload the complex to improve (PDB or mmCIF), or fetch one from the PDB by its ID. The designed chain must be a standard protein with complete backbone atoms; the other chains are its partners and stay as they are. The first partner chain is the one the interface is scored against. Ions and ligands may sit in a chain; each one is read as an entity of its own. Up to 500 residues and ligand atoms in all.
Written onto the chain's backbone before the first cycle, for example an mpnn-design sequence for a generated backbone. It needs one letter per residue (at most 400). Leave it empty to start from the residues in the file.
An A3M or FASTA alignment for a partner protein chain, used by both predictors. Without one the chain is predicted from its sequence alone. The designed chain takes none: its sequence changes every cycle.
Each cycle is one predictor run and one LigandMPNN pass. The cycles and the search steps are capped together (about 4200 s of work per reference GPU).
1 to 20.
0 to 1: the chance that a cycle uses the RoseTTAFold3 sequence search instead of Boltz-2. The choice follows the seed.
1 to 20.
1 to 3. The first sample is scored and carried forward.
0 to 100. Designed-chain runs of five residues or more above this stay unchanged in the redesign.
0.01 to 1. Lower values give more conservative sequences.
Set it to reproduce a run; leave it empty for the default seed.
The confidence score the sequence search minimizes.
One sample per cycle is carried forward.