Rfd4 Proteina Tools#

Configuration File: remote_tools/rfd4_proteina_tools.json Tool Type: Remote Tools Count: 1

This page contains all tools defined in the rfd4_proteina_tools.json configuration file.

Available Tools#

rfd4_proteina_design (Type: RemoteTool)#

Design biomolecules with RFD4-Proteina. The input is an annotated mmCIF whose annotations ARE the…

rfd4_proteina_design tool specification

Tool Information:

  • Name: rfd4_proteina_design

  • Type: RemoteTool

  • Description: Design biomolecules with RFD4-Proteina. The input is an annotated mmCIF whose annotations ARE the design specification: what to hold fixed by coordinate and by sequence, which residues are hotspots, which regions to generate. One call covers every task upstream serves – de novo monomers, protein and DNA binders, motif scaffolding, enzyme active-site scaffolding, small-molecule binders, antibody and TCR CDR redesign, partial diffusion and inverse folding – because which one you get is a property of the annotations, not of a mode. Returns each design’s per-chain sequence; the mmCIF text itself is opt-in.

Parameters:

  • input_cif (string) (required) The annotated mmCIF text. An ordinary unannotated structure is not a design specification and will be refused. Upstream’s examples/ directory has one per design shape.

  • num_replicates (integer) (optional) Independent designs from the one specification, 1 to 8. Each holds the GPU for a whole rollout, so this multiplies the wall clock.

  • recycle_strategy (string) (optional) Compute preset. The three differ only in how often the trunk is recycled over 60 steps: thinking_low 120 trunk passes, thinking_medium 180, thinking_high 300.

  • inference_backend (string) (optional) ‘native’ runs the workflow in torch; ‘bioir’ runs it through NVIDIA’s BioNeMo Inference Runtime, which captures a CUDA graph. See this provider’s README for what each measured on the verified card before choosing.

  • seed (integer) (optional) Random seed. Omit for a random one.

  • return_structures (boolean) (optional) Return the designed mmCIF text as well as the sequences. Off by default: replicates of mmCIF are megabytes, and a transport truncates an oversized reply into invalid JSON rather than refusing it.

Example Usage:

query = {
    "name": "rfd4_proteina_design",
    "arguments": {
        "input_cif": "example_value"
    }
}
result = tu.run(query)