Sequence design conditioned on ligand atomic context (LigandMPNN)

LigandMPNN is an inverse folding model designed for protein-ligand/nucleic acid/metal atomic context-conditioned sequence design. Unlike methods that design sequences based solely on protein backbone, it simultaneously considers the spatial distribution of surrounding non-protein atoms, making it more suitable for sequence optimization of binding pockets, catalytic sites, metal-binding sites, and nucleic acid-binding interfaces.

After uploading a PDB or mmCIF structure file, the system will automatically parse chain information. Select chains to design, configure context parameters and constraints, and the model will generate multiple candidate sequences, returning the score, global score, and sequence recovery for each.

1. Upload Protein Structure File (PDB / mmCIF):

3. Design Parameters:

Model:
Num. Sequences:		Generate 1-10 candidate sequences per run
Sampling Temp:		Higher temperature = greater diversity
Random Seed:		0 = random each time
Use Atom Context:		When checked, considers ligands, nucleic acids, metals, and other non-protein atoms
Binding Site Score Cutoff:		A; defines the neighborhood range for ligand confidence
Use Fixed Residue Sidechains as Context:		No sidechain packing; uses existing fixed residue sidechains as additional atomic context only
Parse Zero-Occupancy Atoms:		By default, atoms with occupancy=0 are not parsed
Exclude AAs:		Checked amino acids will not appear in generated sequences

4. Constraints & Biases (optional, residue numbering is PDB-based):

Fixed residues (space-separated):	Redesign only these residues (space-separated):
Global AA bias (bias_AA):	Per-residue AA bias JSON:
Per-residue AA exclusion JSON:	Parse only these chains (optional, comma-separated):

About LigandMPNN

LigandMPNN is an atomic context-conditioned protein sequence design model.

Key features:
  - Takes protein backbone coordinates as input while simultaneously considering surrounding non-protein atoms
  - Suitable for structural design involving small molecules, metal ions, nucleic acids, and other contexts
  - Supports chain-level design / context chain fixation
  - Supports fixed residues, local redesign, AA bias, per-site AA exclusion
  - The current online version provides sequence design only, without sidechain packing and PDB output

Recommendations:
  - If functional ligands or metals are present near the target site, keep “Use Atom Context" enabled
  - If you only want to design using the protein backbone, you can disable atom context
  - If the input structure contains important fixed residues and you wish to leverage their sidechain geometry, enable “Use Fixed Residue Sidechains as Context"

References

Dauparas J, Lee GR, Pecoraro R, An L, Anishchenko I, Glasscock C, Baker D. Atomic context-conditioned protein sequence design using LigandMPNN. Nat Methods. 2025 Apr;22(4):717-723. doi: 10.1038/s41592-025-02626-1. Epub 2025 Mar 28. PMID: 40155723; PMCID: PMC11978504.