Tools · Computational Biology

OpenMM MD

Prepare a protein system and run a short Langevin equilibration with OpenMM, with optional RMSF, contact, buried-surface, and distance analytics.

Updated October 1, 2026

On this page

Prices, workflows, and method papers Open in the app

OpenMM MD takes a protein structure from PDB to an equilibrated simulation in one job: it cleans the structure, assigns protonation states, solvates it, minimizes, and runs Langevin dynamics with OpenMM, then optionally analyzes the trajectory. It runs on a GPU by default (T4); CPU is for short test runs.

How it works

Prepare. PDBFixer repairs missing atoms and residues (and removes heterogens if you ask), PROPKA assigns protonation states at preparation.ph (7.4 by default), and OpenMM's Modeller adds a water box with padding_nm (1.0) of padding and ions at ionic_strength (0.15 M).
Simulate. Energy minimization, then simulation.num_steps of Langevin dynamics (up to 500,000) at temperature_k (300 K). Hydrogen mass repartitioning (hmr, on by default) allows a 4 fs time step instead of 2 fs, so the default 50,000 steps cover about 0.2 ns.
Analyze (optional). Each entry in analytics runs on the trajectory with MDAnalysis and writes a CSV.

Ensembles

`simulation.ensemble`	Behavior
`nvt` (default)	Constant volume
`npt`	Adds a Monte Carlo barostat at `pressure_bar` (1.0 by default). Needs explicit water.

Force fields and solvent

`preparation.force_field`	`preparation.solvent_model`
`amber14` (default)	`tip3pfb` (default), `tip3p`, `tip4pew`, `spce`, or `implicit_obc2`
`amber99sbildn`	`tip3p` or `implicit_obc2`

implicit_obc2 skips the water box and uses OBC2 generalized-Born implicit solvent, with NVT only. Only force fields we can license for this use are offered; CHARMM36 for proteins isn't available yet.

Analytics

`metric`	Fields	Measures
`rmsf`	`selection`	Per-atom fluctuation over the trajectory
`contact_map`	`selection_a`, `selection_b`, `cutoff` (4.5 Å)	How often each pair of residues is in contact
`buried_surface_area`	`selection_a`, `selection_b`	Surface area buried at the interface
`com_distance`	`selection_a`, `selection_b`	Distance between the two centers of mass

Selections use MDAnalysis syntax, for example protein and name CA or segid A.

System modes

protein_protein (default): a protein or complex from structure.pdb_text, as above.
membrane_small_molecule: a small molecule from ligand.smiles, parameterized with OpenFF Sage and placed in the water above a small prebuilt lipid bilayer patch (box_id: lnp_hspc_chol_v1, rebuilt as a POPC bilayer for CHARMM36) with 150 mM NaCl. It doesn't need a PDB, but in a workflow the step still needs its structure input connected.

Outputs

File	Contents
`simulation/minimized.pdb`	The minimized starting system
`simulation/trajectory.dcd`	The trajectory
`simulation/equil_log.csv`	Step, potential energy, temperature, and density (plus volume for NPT)
`analytics/<metric>_<n>.csv`	One file per analytics entry
`pipeline_summary.json`	The settings used, files, and analytics results

Intermediate structures (cleaned, protonated, and solvated) are in intermediate/. Analyze the trajectory further with the MD Trajectory Analyzer, or rescore binding with MM-GBSA Rescoring.

Not covered

Constant-energy (NVE) runs.
CHARMM36, COMPASS, and DREIDING for proteins, and implicit solvents other than OBC2.

Run it from the API

Submit with Submit a job and the job_type below. Price it first with Estimate job reservation cost: submitting reserves that amount from your wallet, and the charge settles at the actual runtime.

OpenMM MD `openmm-md`

Job type: openmm-md
Hardware: cput4 (default)l4a10l40sa100-40gba100-80gbh100h200b200
Typical runtime: 2 h on Nvidia T4

Payload

Payload fields
Field	Type	Description
`system_mode`	`string`	V1 accepts protein_protein and membrane_small_molecule at submit time. Default: `"protein_protein"`One of: `"protein_protein", "protein_ligand", "small_molecule", "membrane_small_molecule"`
`structure`required	`object`
structure.`pdb_text`required	`string`	Raw PDB text with at least one ATOM/HETATM record. Required for protein_protein; not required for membrane_small_molecule. Limits: `min length 1`
`preparation`	`object`
preparation.`ph`	`number`	Default: `7.4`Limits: `≥ 0, ≤ 14`
preparation.`remove_heterogens`	`boolean`	Default: `false`
preparation.`padding_nm`	`number`	Default: `1`Limits: `> 0`
preparation.`ionic_strength`	`number`	Default: `0.15`Limits: `≥ 0`
preparation.`force_field`	`string`	Default: `"amber14"`One of: `"amber14", "amber99sbildn"`
preparation.`solvent_model`	`string`	Default: `"tip3pfb"`One of: `"tip3pfb", "tip3p", "tip4pew", "spce", "implicit_obc2"`
preparation.`box_id`	`string`	Prebuilt membrane box id. Defaulted for membrane_small_molecule. Default: `"lnp_hspc_chol_v1"`
`simulation`	`object`
simulation.`num_steps`	`integer`	Default: `50000`Limits: `≥ 1, ≤ 500000`
simulation.`temperature_k`	`number`	Default: `300`Limits: `> 0`
simulation.`hmr`	`boolean`	Default: `true`
simulation.`ensemble`	`string`	Default: `"nvt"`One of: `"nvt", "npt"`
simulation.`pressure_bar`	`number`	Default: `1`Limits: `> 0`
`analytics[]`	`object[]`	Default: `[]`
analytics.`metric`required	`string`	One of: `"rmsf", "contact_map", "buried_surface_area", "com_distance"`
analytics.`selection`	`string`	Limits: `min length 1`
analytics.`selection_a`	`string`	Limits: `min length 1`
analytics.`selection_b`	`string`	Limits: `min length 1`
analytics.`cutoff`	`number`	Limits: `> 0`
`ligand`	`object`
ligand.`smiles`required	`string`	Ligand SMILES. Required for membrane_small_molecule. Limits: `min length 1`

Example

from cognichem_client import CogniChem

client = CogniChem.from_env()  # reads COGNICHEM_API_KEY
payload = {
    "system_mode": "protein_protein",
    "structure": {
        "pdb_text": "REMARK  Minimal alanine tripeptide for unit testing.\nREMARK … (2,166 characters)",
    },
    "simulation": {"num_steps": 1},
    "analytics": [],
}

estimate = client.jobs.estimate(job_type="openmm-md", payload=payload, resource="t4")
print(f"Reserves ${estimate.cost:.2f}")

job = client.jobs.submit(
    job_name="my-openmm-md-run",
    job_type="openmm-md",
    payload=payload,
    resource="t4",
)
status = client.jobs.wait(job.process_id)
if status.status == "completed":
    client.jobs.result(job.process_id, save_path=".")

curl -X POST "https://api.cognichem.com/api/v1/jobs/submit" \
  -H "X-Api-Key: $COGNICHEM_API_KEY" \
  -H "Idempotency-Key: $(uuidgen)" \
  -H "Content-Type: application/json" \
  -d '{
  "job_name": "my-openmm-md-run",
  "job_type": "openmm-md",
  "payload": {
    "system_mode": "protein_protein",
    "structure": {
      "pdb_text": "REMARK  Minimal alanine tripeptide for unit testing.\nREMARK … (2,166 characters)"
    },
    "simulation": {
      "num_steps": 1
    },
    "analytics": []
  },
  "resource": "t4"
}'

{
  "system_mode": "protein_protein",
  "structure": {
    "pdb_text": "REMARK  Minimal alanine tripeptide for unit testing.\nREMARK … (2,166 characters)"
  },
  "simulation": {
    "num_steps": 1
  },
  "analytics": []
}

Sample data from the job catalog; long values are shortened here. Each job_name must be unique among your jobs.

Workflow inputs

Molecules · optionalSMILES
Protein structurePDB

Workflow outputs

ArchiveZIP
Protein structurePDB

OpenMM MD openmm-md

OpenMM MD `openmm-md`