Rfdiffusion2 Tools#

Configuration File: remote_tools/rfdiffusion2_tools.json Tool Type: Remote Tools Count: 1

This page contains all tools defined in the rfdiffusion2_tools.json configuration file.

Available Tools#

rfdiffusion2_design (Type: RemoteTool)#

Design protein backbones with RFdiffusion2, including atom-level enzyme active-site scaffolding a…

rfdiffusion2_design tool specification

Tool Information:

  • Name: rfdiffusion2_design

  • Type: RemoteTool

  • Description: Design protein backbones with RFdiffusion2, including atom-level enzyme active-site scaffolding around a small molecule. Give a contig string and a provider-side PDB that holds the motif residues, the ligands, and RFdiffusion2’s ORI positioning atom. Returns each designed backbone as PDB text with residue, ligand, and motif summaries. One 180-residue design takes about 3.5 minutes on a 32 GB GPU.

Parameters:

  • contig_map (string) (required) RFdiffusion2 contig string. Numbers are lengths of new backbone; chain-prefixed ranges are motif residues taken from input_pdb. Example: ‘46,A106-106,59,A166-166,2,A169-169,23,A193-193,46’.

  • input_pdb (string) (required) PDB file name relative to the provider’s RFdiffusion2 input directory, for example ‘mcsa_41/M0584_1ldm.pdb’. The file must hold the motif residues, the ligands, and an ORI HETATM atom marking where the design is centered. Absolute paths, ‘..’ segments, and URLs are rejected.

  • ligand (string) (optional) Comma-separated residue names of the ligands in input_pdb to design around, for example ‘NAD,OXM’.

  • contig_atoms (object) (optional) Atom-level motif: maps a motif residue such as ‘A106’ to the comma-separated atom names to keep, for example ‘NE,CD,CZ’.

  • contig_as_guidepost (boolean) (optional) If true, RFdiffusion2 chooses where in the sequence the motif residues go instead of fixing them to the contig order.

  • num_designs (integer) (optional) Number of backbones to generate.

  • inference_steps (integer) (optional) Number of flow-matching steps. Omit to use the model default of 100.

  • seed (integer) (optional) Seed for a reproducible run. Omit for a random design.

  • return_structure (boolean) (optional) If false, return only the per-design summaries without the PDB text.

  • timeout_seconds (integer) (optional) Maximum time for the call in seconds, including any wait for another design to finish on the provider.

  • dry_run (boolean) (optional) If true, validate the arguments, including that input_pdb exists on the provider, and return the Hydra overrides without running RFdiffusion2.

Example Usage:

query = {
    "name": "rfdiffusion2_design",
    "arguments": {
        "contig_map": "example_value",
        "input_pdb": "example_value"
    }
}
result = tu.run(query)