RF Diffusion Scaffold Guided Config

Builds a scaffold-guided (fold-conditioned) configuration for RF Diffusion protein design. It bundles scaffold PDBs, optional target protein inputs for binder design, loop-masking behavior, residue sampling at termini/loops, and optional contig cropping into a single configuration object.

Usage

Use this node when you want RF Diffusion to follow a specific fold/topology using scaffold structures. Connect scaffold PDBs (from a PDB Loader or PDB Combiner). For protein-protein interaction design, also supply a target PDB and the target PDBs used to derive secondary structure. Feed the resulting configuration into the RF Diffusion node’s scaffoldguided_config input.

Inputs

Field	Required	Type	Description	Example
scaffold_pdbs	True	PDB	Dictionary of scaffold PDB structure(s) used to impose fold conditioning. Keys are names; values are PDB file contents.	{'scaffold_1.pdb': '', 'scaffold_2.pdb': ''}
target_pdb	False	PDB	Single target protein PDB for conditioned binder design. When provided, it is recommended to also provide target_pdbs for computing target secondary structure.	{'target.pdb': ''}
target_pdbs	False	PDB	Dictionary of target PDB structure(s) used to compute secondary structure for the target (recommended when target_pdb is set).	{'target_ref_1.pdb': '', 'target_ref_2.pdb': ''}
scaffold_list	False	STRING	Content of a scaffold list (one scaffold name per line). When provided, only listed scaffolds are used. Lines starting with '#' are treated as comments.	scaffold_1.pdb scaffold_2.pdb # comment line
sampled_insertion	False	INT	Maximum number of residues to insert into loops per run to extend or adjust secondary structures. Applied only if > 0.	5
sampled_N	False	INT	Maximum number of additional residues to sample at the N terminus. Applied only if > 0.	3
sampled_C	False	INT	Maximum number of additional residues to sample at the C terminus. Applied only if > 0.	2
ss_mask	False	INT	Number of residues to mask at the end of each secondary structure block (0–20). Applied only if > 0.	4
systematic	False	BOOLEAN	If true, iterates through scaffolds systematically rather than randomly sampling.	True
mask_loops	False	BOOLEAN	Whether to mask loops. True allows extra residues without precise structures; typically set False for PPI designs.	False
contig_crop	False	STRING	Residue/chain crop spec to select a subset from the scaffold as the starting template. Uses contig-like syntax (e.g., 'A10-50', optionally multiple segments).	A10-50;B5-30

Outputs

Field	Type	Description	Example
scaffoldguided_config	RFDIFFUSION_SCAFFOLDGUIDED_CONFIG	A configuration object enabling scaffold-guided mode with secondary structure/adjacency inputs, optional target conditioning, sampling and masking parameters. Connect this to the RF Diffusion node’s scaffoldguided_config input.	{'scaffoldguided': True, 'scaffold_secstruc_adj': {'type': 'secstruc_adj_input', 'pdbs': {'scaffold_1.pdb': ''}}, 'mask_loops': False, 'sampled_insertion': 5, 'sampled_N': 3, 'sampled_C': 2, 'ss_mask': 4, 'systematic': True, 'contig_crop': 'A10-50', 'target_pdb': True, 'target_pdb_content': '', 'target_secstruc_adj': {'type': 'secstruc_adj_input', 'pdbs': {'target_ref_1.pdb': ''}}}

Important Notes

• Provide scaffold_pdbs as a dictionary mapping names to PDB contents; at least one scaffold is required.
• For binder (PPI) designs, supply target_pdb and, ideally, target_pdbs to compute target secondary structure.
• Loop and termini sampling parameters (sampled_insertion, sampled_N, sampled_C) are applied only when greater than zero.
• Set mask_loops to False for typical PPI workflows to avoid masking loops on the target interface.
• contig_crop must follow contig-like residue selection syntax (e.g., A10-50). Invalid specs will be rejected.
• scaffold_list filters which scaffolds are used; ensure listed names match the keys in scaffold_pdbs.

Troubleshooting

• Error: 'scaffold_pdbs is required' — Ensure scaffold_pdbs is connected and is a non-empty dictionary of name->PDB content.
• Warning: target_pdb provided without target_pdbs — The config can still run, but provide target_pdbs for better target secondary structure derivation.
• No effect from sampling parameters — Confirm sampled_insertion, sampled_N, or sampled_C are set to values > 0.
• Invalid contig_crop format — Use chainID followed by residue ranges (e.g., 'A10-50'). Remove spaces and ensure chain IDs exist in the scaffold.
• Scaffolds not filtered by scaffold_list — Verify each line matches a key in scaffold_pdbs and that comment lines start with '#' only.