Af2Ig Tools#

Configuration File: remote_tools/af2ig_tools.json Tool Type: Remote Tools Count: 1

This page contains all tools defined in the af2ig_tools.json configuration file.

Available Tools#

af2ig_predict (Type: RemoteTool)#

Re-predict a designed binder-target complex with AlphaFold 2 using the complex itself as the stru…

af2ig_predict tool specification

Tool Information:

  • Name: af2ig_predict

  • Type: RemoteTool

  • Description: Re-predict a designed binder-target complex with AlphaFold 2 using the complex itself as the structural initial guess (nrbennet/dl_binder_design af2_initial_guess, model_1_ptm). This is the standard in-silico filter for designed binders: give it the complex as PDB text with the BINDER AS THE FIRST CHAIN and the target as the chains after it, and it returns the predicted complex and the interface confidence scores. The headline number is pae_interaction, in angstroms, where the field’s working threshold is below 10. No MSA is built, no template is searched and no external service is contacted; the result is deterministic. The input must be a structure – there is no sequence-only route.

Parameters:

  • structure (string) (required) The complex as PDB text. CHAIN 1 IS THE BINDER and every following chain is the target; a single-chain file is predicted as a monomer and the interface scores come back null. ATOM records of standard residues are read; HETATM, waters and ligands are ignored. Backbone-only coordinates are enough.

  • num_recycle (integer) (optional) AlphaFold recycling iterations (upstream’s default is 3). This is compile-time: changing it rebuilds the held model and recompiles, which the reply reports as config_rebuilt.

  • initial_guess (boolean) (optional) Use the input coordinates as the structural initial guess (default true, and the whole point of the method). False runs the same template-conditioned prediction without it, which is a different and weaker filter. Compile-time, as num_recycle is.

  • force_monomer (boolean) (optional) Treat the first chain alone as the whole input and drop the rest. Used to score a binder’s own fold independently of its target.

  • max_amide_dist (number) (optional) Angstroms. A C-N amide distance above this is read as a chain break and gets AlphaFold’s +200 residue-index offset (upstream’s default 3.0).

  • return_structure (boolean) (optional) Return the predicted complex as PDB text, with per-residue pLDDT in the B-factor column (default false: the scores are what a filter reads, and a few-hundred-residue PDB is hundreds of kilobytes). This is the ONLY way to get the structure: the server keeps no output directory and returns no host path.

Example Usage:

query = {
    "name": "af2ig_predict",
    "arguments": {
        "structure": "example_value"
    }
}
result = tu.run(query)