Bioemu2 Tools#

Configuration File: remote_tools/bioemu2_tools.json Tool Type: Remote Tools Count: 1

This page contains all tools defined in the bioemu2_tools.json configuration file.

Available Tools#

bioemu2_sample_structures (Type: RemoteTool)#

Sample all-atom 3D structures of a biomolecular system with BioEmu-2. Give it the composition – …

bioemu2_sample_structures tool specification

Tool Information:

  • Name: bioemu2_sample_structures

  • Type: RemoteTool

  • Description: Sample all-atom 3D structures of a biomolecular system with BioEmu-2. Give it the composition – protein chains as sequences, or a system definition that may also contain ligands, RNA, DNA, modified residues and non-standard bonds – and it returns an ensemble of conformations, not a single structure. ‘structure’ mode asks for one accurate fold; ‘equilibrium’ mode asks for a spread of conformations from the modelled equilibrium ensemble. Sampling frequency is NOT a calibrated Boltzmann probability and the model predicts neither sequences nor binding affinities. Structures are written to the worker’s output directory; per-frame physicality checks come back either way. The system must be at most 448 tokens, which is BioEmu-2’s own count and includes a ligand’s atoms, so it is larger than the residue count; a bigger system is refused by name.

Parameters:

  • sequences (array) (optional) One one-letter amino-acid sequence per protein chain, lettered A, B, C… in the order given. Identical sequences become one entity with several chains. Use ‘system_yaml’ instead for anything that is not a plain protein chain.

  • system_yaml (string) (optional) BioEmu-2’s own system YAML, as written by System.to_yaml(): an ‘entities’ list and a ‘bonds’ list. This is the way to describe ligands by CCD code or SMILES, nucleic acids, modified residues, explicit chain ids and covalent bonds. Mutually exclusive with ‘sequences’.

  • mode (string) (optional) ‘structure’ sharpens towards one accurate fold; ‘equilibrium’ samples the modelled ensemble without sharpening. They differ in the sampler, not in what they accept.

  • num_samples (integer) (optional) Conformations to generate, 1 to {…}. On the verified card this barely moved peak memory at 76 residues but it does cost wall-clock.

  • msa_a3m (object) (optional) Optional alignment per chain id, as A3M text whose first record is that chain’s query sequence. OMITTING THIS SENDS A QUERY TO THE PUBLIC COLABFOLD SERVER on a cache miss, which leaves this machine and depends on a third party. Supply your own to stay local; a query-only alignment is valid A3M and degrades the model rather than failing.

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

  • physicality (string) (optional) Which geometry checks to report per frame: chain breaks, clashes and bond lengths. Failing frames are still returned and still saved; this only labels them.

  • return_topology (boolean) (optional) Return the topology.cif text in the reply as well as writing it. Off by default: a transport truncates an oversized reply into invalid JSON rather than refusing it. The trajectory frames are never inlined.

Example Usage:

query = {
    "name": "bioemu2_sample_structures",
    "arguments": {
    }
}
result = tu.run(query)