Complexa Tools#

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

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

Available Tools#

complexa_design_binder (Type: RemoteTool)#

Design de novo protein binders against a target with Proteina-Complexa 1.1.0 (NVIDIA’s flow-match…

complexa_design_binder tool specification

Tool Information:

  • Name: complexa_design_binder

  • Type: RemoteTool

  • Description: Design de novo protein binders against a target with Proteina-Complexa 1.1.0 (NVIDIA’s flow-matching binder generator). Give the target structure as PDB text, the chain and residue range to crop it to (e.g. ‘A1-115’), and hotspot residues on its surface (e.g. ‘A37’); it samples binders of the requested length and returns, per design, the decoded binder amino-acid sequence, with the full complex (chain A target crop, chain B new binder) as PDB text on request. Generation stage only: samples are raw, not scored or ranked.

Parameters:

  • target_pdb (string) (required) The target protein structure as PDB text (ATOM records). Residue numbers in it are the ones target_input and hotspot_residues refer to.

  • target_input (string) (required) Chain and residue range of the target to design against, e.g. ‘A1-115’ (at most 300 residues).

  • hotspot_residues (array) (required) Target residues the binder should contact, each <chain><number> and inside target_input, e.g. [‘A37’, ‘A39’, ‘A49’].

  • binder_length (integer) (optional) Residues in each designed binder, 20..200 (default 80).

  • num_designs (integer) (optional) Designs to sample, 1..16 (default 1).

  • batch_size (integer) (optional) Designs per forward pass (default: all of them). Lowering it trades time for peak GPU memory; (target crop + binder)^2 x batch_size is bounded at 4000000.

  • num_steps (integer) (optional) Flow-matching integration steps per design (10..400, default 400, upstream’s own). Fewer is faster and rougher.

  • seed (integer) (optional) Random seed (upstream’s default is 5). The same seed and inputs reproduce the designs.

  • name (string) (optional) A label for this target; defaults to ‘design’.

  • return_pdb (boolean) (optional) Return each design’s PDB text (default false: ~100 KB per design). Sequences come back either way.

Example Usage:

query = {
    "name": "complexa_design_binder",
    "arguments": {
        "target_pdb": "example_value",
        "target_input": "example_value",
        "hotspot_residues": ["item1", "item2"]
    }
}
result = tu.run(query)