Ef2Inv Tools#

Configuration File: remote_tools/ef2inv_tools.json Tool Type: Remote Tools Count: 1

This page contains all tools defined in the ef2inv_tools.json configuration file.

Available Tools#

ef2inv_design (Type: RemoteTool)#

Design a binder SEQUENCE against a target protein with the ESM cookbook’s gradient-guided ESMFold…

ef2inv_design tool specification

Tool Information:

  • Name: ef2inv_design

  • Type: RemoteTool

  • Description: Design a binder SEQUENCE against a target protein with the ESM cookbook’s gradient-guided ESMFold2 recipe. Give a built-in target name (cd45, ctla4, egfr, pd-l1, pdgfr) or any label plus its ‘target_sequence’, and a ‘binder_len’. Returns each design’s binder sequence and the four hero critics’ scores; the refolded structure is opt-in. One design is a fixed 150-step optimisation – the step count is the recipe’s own constant and cannot be lowered.

Parameters:

  • target_name (string) (required) A built-in target (cd45, ctla4, egfr, pd-l1, pdgfr: the recipe’s own sequences, give no target_sequence), or any label for a target_sequence you supply.

  • target_sequence (string) (optional) The target’s amino-acid sequence, required when target_name is not one of the built-ins and forbidden when it is. Up to 280 residues.

  • binder_len (integer) (optional) Length of the binder to design, 8-200. target_len + binder_len must not exceed 300 tokens.

  • n_designs (integer) (optional) How many designs to run (default 1, max 4). Each is an independent 150-step optimisation at seed+k, run one after another: the count costs time, not memory.

  • seed (integer) (optional) Random seed of the first design.

  • target_hotspot_ids (array) (optional) Target residue ids to treat as the epitope, e.g. [“56”, “57”]. Omit to let the recipe choose.

  • epitope_contact_distance (number) (optional) The recipe’s epitope contact distance in angstroms.

  • warm_only (boolean) (optional) Run the recipe’s one-time costs and a single design-step-shaped gradient pass instead of a design (seconds rather than minutes). Use it to check that the provider is live and what it costs on this card.

  • return_structures (boolean) (optional) Return each design’s refolded mmCIF text (default false: hundreds of kilobytes per design).

Example Usage:

query = {
    "name": "ef2inv_design",
    "arguments": {
        "target_name": "example_value"
    }
}
result = tu.run(query)