Rosetta Fast Relax¶
Runs Rosetta FastRelax to refine protein structures. Given one or more sequences (FASTA) and corresponding initial structures (PDB), it performs a specified number of relaxation cycles to improve local geometry. Returns relaxed PDB structures, preserving input-to-output mapping by record IDs.
Usage¶
Use this node after generating initial protein structures (e.g., from a folding model) to improve stereochemistry and relieve clashes. Provide the sequences and corresponding PDBs with matching IDs. Increase relax_cycles for more thorough refinement at the cost of runtime.
Inputs¶
| Field | Required | Type | Description | Example |
|---|---|---|---|---|
| fasta | True | FASTA | Protein sequence(s) in FASTA format. If multiple sequences are provided, each record ID must match the corresponding structure ID in the PDB input. | >protein1 MKTAYIAKQRQISFVKSHFSRQ... >protein2 GHHHHHHSSGLVPRGSHMASMT... |
| pdb | True | PDB | Initial structure(s) to relax. Provide as a dictionary mapping IDs to PDB text. IDs must correspond to the FASTA record IDs. | {"protein1": "ATOM 1 N MET A 1 ...\nEND\n", "protein2": "ATOM 1 N GLY A 1 ...\nEND\n"} |
| relax_cycles | True | INT | Number of relaxation cycles to apply to each structure. Higher values can yield better local geometry but increase runtime. | 3 |
| debug | True | BOOLEAN | If true, any pose error stops the run with an error. If false, the run continues and skips poses that fail. | true |
Outputs¶
| Field | Type | Description | Example |
|---|---|---|---|
| structure.pdb | PDB | Relaxed structures as a dictionary mapping input IDs to PDB content. | {"protein1": "ATOM ...\nEND\n", "protein2": "ATOM ...\nEND\n"} |
Important Notes¶
- ID matching required: FASTA record IDs must correspond exactly to the structure IDs in the PDB input.
- Not deterministic: A seed parameter is not exposed; results may not be bitwise reproducible across runs.
- Runtime scales with size: Runtime increases with the number of sequences and relax_cycles.
- Error handling: With debug=true, any failure raises an error. With debug=false, failures for some poses are skipped; only successful relaxations are returned.
- Purpose: This node refines geometry; it does not perform folding from sequence.
- Batch support: Multiple sequences/structures can be processed in one run, provided IDs align.
Troubleshooting¶
- Missing outputs for some IDs: If debug=false, failed poses are skipped. Re-run with debug=true to see the exact error, or reduce relax_cycles.
- ID mismatch error: Ensure FASTA headers (record IDs) match keys in the PDB dictionary exactly (including capitalization).
- Long runtimes: Lower relax_cycles or process fewer sequences per run. Check that input structures are reasonable (severely distorted inputs relax more slowly).
- Invalid FASTA/PDB formats: Validate that FASTA is properly formatted and each PDB entry contains standard ATOM/HETATM records with chain/residue numbering.
- Downstream compatibility: If downstream tools expect specific residue numbering, verify that relaxation preserved the mapping you rely on.