BioEmu2

Overview

Equilibrium-ensemble sampling of protein flexibility and conformational states.
Structure prediction with sharpened sampling.
Time-lagged propagation of starting structures forward in time (advanced mode).
Systems can include proteins, nucleic acids, ligands, and modified residues via system YAML.

Modes

Mode	Input shape	When to use it
`sampling` Sample Ensemble Default	Consumes one A3M alignment file.	Generate a conformational ensemble or predicted structures from an alignment. The A3M query row defines the protein system and the MSA.

Canonical Job Configuration

These are the fields exposed by the default job configuration for bioemu2. They are also returned by GET /api/v1/program/params?program=bioemu2 and submitted as the params JSON object to POST /api/v1/job/submit.

Parameter	Type	What it does
`alignment_file` A3M Alignment	File	A3M alignment whose first record is the query sequence; it defines the protein system and the MSA. Required
`task` Task	Text	Equilibrium ensemble samples many conformations; structure prediction sharpens sampling toward an accurate folded structure. Default: Equilibrium ensemble; Options: Equilibrium ensemble, Structure prediction
`sample_count` Samples	Integer	Number of conformations to generate. More samples increase runtime. Default: 100; Range: 1-10000
`sampler` Sampler	Text	EDM is unsteered; the guided presets steer sampling with physicality potentials and cost more runtime. Default: edm; Options: edm, edm_gd, edm_fk, edm_fkc
`physicality_filter` Physicality Filter	Text	Per-frame physicality checks recorded in physicality.json; frames are always kept in the trajectory. Default: high_quality; Options: high_quality, high_quality_per_residue, no_violations, no_filter
`random_seed` Random Seed	Integer	Set for reproducible sampling; leave blank for a random seed. Range: 1-999999999

Outputs

samples.xtc trajectory with all generated conformations.
topology.cif matching topology; also published as a canonical structure artifact.
physicality.json per-frame status (passed, rejected, or not_checked) and first failed check.
system.yaml resolved system and sampling mode.

The Results view summarizes sample counts, the physicality pass rate, and a per-frame status table; the topology opens in the structure previewer.

Common Examples

Equilibrium ensemble: upload an A3M alignment, 100 samples, Equilibrium ensemble task.
Folded-structure hypothesis: same alignment with the Structure prediction task and a small sample count.
Reproducible run: set a random seed.

Example API params

{
  "alignment_file": "trpcage.a3m",
  "task": "Equilibrium ensemble",
  "sample_count": 100,
  "random_seed": 42
}

Caveats

Jobs run fully offline: ColabFold MSA queries, RCSB downloads, and HuggingFace checkpoint downloads are disabled, so an A3M alignment is always required.
Rejected frames remain in samples.xtc; use physicality.json or the Results table to identify them.
Ensembles are model-generated hypotheses, not free-energy measurements.

Advanced Submit

Advanced submit is still available for direct program arguments through POST /api/v1/job/submit-advanced. Prefer canonical configuration unless you need exact low-level arguments or are reproducing a known command line.

sample supports --a3m, --num-samples, --mode, --sampler, --seed, --batch-size, --num-steps, --physicality-filter, --precision, --matmul-precision, and --system for multi-chain or ligand systems.
propagate adds --x0-cif, exactly one of --x0-xtc or --num-samples, and --lag-time-ns.

curl -X POST https://subseq.bio/api/v1/job/submit \
  -H "Authorization: Bearer <api_key>" \
  -F program=bioemu2 \
  -F 'params={"alignment_file":"trpcage.a3m","task":"Equilibrium ensemble","sample_count":100}'