ProteinMPNN

Design protein sequences for a supplied backbone structure.

Runs the official ProteinMPNN fixed-backbone sequence-design script on selected PDB chains.

Source code ↗

Input

Configure a small job

POST

Job name

Backbone mmCIF or PDB file mmCIF is converted to legacy PDB before design; chain ids in it must be a single character.

Chains to design Comma- or space-separated PDB chain IDs.

Designed residues (optional) Restrict design to specific residues instead of a whole chain: one line per chain, e.g. "B 26 27 28 29 30". Chains left out are designed in full; residues not listed for a named chain are held fixed.

Tie designed chains as a homo-oligomer Symmetric copies: ties the same position across all designed chains to the same amino acid.

Model type SolubleMPNN is trained on soluble proteins only, for designs meant to express without a membrane or aggregation-prone context.

Model

Sequences

Sampling temperature

Noise level Gaussian noise added to backbone coordinates at inference, separate from the model checkpoint's own training noise (v_48_002/010/020/030).

Random seed

Omit amino acids One-letter amino-acid codes that ProteinMPNN must not sample anywhere.

Amino-acid bias (optional) Global per-amino-acid sampling bias, e.g. "W:3.0,P:3.0,C:3.0,A:-3.0".

Per-residue amino-acid bias (optional) JSON object keyed by chain+residue, e.g. {"C1": {"G": -0.3, "P": 10.8}, "C3": {"P": 10.0}}.

Per-residue omitted amino acids (optional) JSON object keyed by chain+residue, e.g. {"A1": "CP", "A2": "CW"}.

Ready

Response

JSON result

{
  "status": "Ready for a request"
}