Atlasfold Tools#

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

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

Available Tools#

atlasfold_predict (Type: RemoteTool)#

Predict an all-atom protein structure with AtlasFold v1.0.0. Give one sequence in ‘chains’ for a …

atlasfold_predict tool specification

Tool Information:

  • Name: atlasfold_predict

  • Type: RemoteTool

  • Description: Predict an all-atom protein structure with AtlasFold v1.0.0. Give one sequence in ‘chains’ for a monomer, or several for one complex (folded by the AtlasFold-M multimer head). Returns per-sample confidence scores (avg_plddt, pTM) and which sample ranked best; the mmCIF text itself is opt-in. Protein only – this model has no ligand, DNA or RNA input.

Parameters:

  • chains (array) (required) The amino-acid sequences of one target. One element folds through the monomer head; two or more are one complex folded through the multimer head.

  • name (string) (optional) A label for the target; appears in the output names.

  • seeds (array) (optional) Random seeds. Each seed produces num_samples structures, so total structures is seeds x num_samples. Upstream’s default is the single seed 1.

  • num_samples (integer) (optional) Diffusion samples per seed (upstream’s default is 5). Each is scored separately, so this is how to compare conformations; rank them on scores.

  • num_recycles (integer) (optional) Recycling iterations (upstream’s default is 4 for a monomer, 10 for a multimer). More costs time, not memory.

  • num_steps (integer) (optional) Diffusion sampling steps. Omit for upstream’s own length-dependent default (20 up to 512 tokens, 30 to 1024, 100 above; 200 for a multimer).

  • format (string) (optional) Structure file format (default cif).

  • return_structures (boolean) (optional) Return the structure text (default false: an all-atom file of a few hundred residues is megabytes). The confidence scores come back either way.

Example Usage:

query = {
    "name": "atlasfold_predict",
    "arguments": {
        "chains": ["item1", "item2"]
    }
}
result = tu.run(query)