What SimAtomic runs.
And what the results mean.

Start with a supported system

Upload the receptor as PDB or CIF, plus an optional separate SDF containing one chemically valid ligand already positioned relative to the receptor. Supply the biological assembly you intend to simulate. SimAtomic does not retrieve database structures, reconstruct assemblies from symmetry, or generate a docking pose.

From structure to trajectory

1. Receptor repair and protonation

CIF is converted to PDB with Open Babel; some mmCIF metadata and long chain identifiers can be lost. PDBFixer adds missing heavy atoms in standard residues. Missing terminal sequence gaps are dropped rather than rebuilt; only limited internal gaps can be modeled. Large missing loops or domains need external preparation.

Crystallographic waters are removed and fresh solvent is added later. Receptor protonation is assigned with pdb2pqr at the configured pH, followed by hydrogen handling during solvation. Existing curated hydrogen and protonation assignments are not guaranteed to survive. Catalytic and metal-binding residues require particular scrutiny.

2. Ligand merge, water, and ions

The ligand SDF is used as supplied; the receptor pH setting does not choose its protonation or tautomer. OpenFF may infer missing hydrogens from valence, which does not establish the intended chemical microstate. The combined solute is oriented and placed in a periodic hexagonal prism with 1.2 nm padding by default, neutralizing ions, and 0.15 M NaCl.

3. Force-field parameterization

The pipeline constructs the molecular topology and OpenMM system, including atom types, partial charges, bonded interactions, and nonbonded forces. Inspect the prepared coordinates and topology before interpreting a trajectory. Solvent clash cleanup does not repair an incorrect ligand pose or chemistry.

4–6. Relaxation, production, and imaging

Restrained energy minimization is followed by heating and equilibration. Production uses Langevin dynamics with NVT as the default and optional NPT pressure control. Periodic imaging wraps the trajectory for inspection and analysis. Backend restart and selected-residue harmonic restraint options are available when exposed by the chosen workflow.

The physical model

Classical MD propagates atomic coordinates using a parameterized potential energy function. The standard path uses fixed-charge, all-atom force fields and explicit solvent. It describes structural motion within that model; it does not model chemical reactions or electronic rearrangement.

HMR can enable a 4 fs integration step with appropriate constraints and validation. The listed base timestep is 2 fs; enabling HMR alone is not evidence that a particular job used 4 fs. Read the effective run configuration and trajectory timing. For the underlying simulation concepts, see the OpenMM user guide and OpenFF force-field documentation.

Parameters and units

This is a reference for the backend configuration, not a promise that every field is accepted by every interface. The MCP server’s current tool schema determines which fields an assistant can submit. Use the API configuration supplied for your deployment and preserve the effective settings returned with your results.

Preparation and production

Production duration: steps × timestep in ps ÷ 1,000 = ns. At 0.002 ps per step, 500,000 steps is 1 ns. Saving every 5,000 steps gives a 10 ps interval. These are simulation times, not GPU wall-clock time or a credit estimate.

The MCP submission tool also describes duration inputs simulation_time_ps and simulation_time_ns. Use one duration representation accepted by the current schema; do not send contradictory duration fields.

Analysis settings

Advanced backend options include heating and equilibration schedules, barostat frequency, harmonic restraints, state loading, trajectory output, and optional AMBER .prmtop export. Their availability and names depend on the exposed API or MCP schema.

Workflows and scientific interpretation

Protein or protein–ligand MD

Inspect structural stability, fluctuations, contacts, and changes in a supplied pose. Start with a short preparation and stability test, then choose production length and independent replicates for the question. A pose that remains stable briefly is not proof of binding affinity.

Trajectory analysis

Analyze downloaded coordinates with local scripts or submit a follow-up analysis job. Time-lagged independent component analysis (TICA) and clustering can summarize sampled conformations. The result depends on atom selection, frame spacing, sampling length, and model settings; clusters do not establish complete conformational sampling.

Optional MM/PBSA and MM/GBSA

The optional end-point workflow and AMBER topology export currently target protein–ligand systems. These calculations estimate contributions from molecular mechanics and continuum solvation over selected snapshots. An igb setting selects a Generalized Born model, so that calculation is MM/GBSA rather than a Poisson–Boltzmann calculation. Treat the output as an approximate, model-dependent comparison, not a rigorous binding free energy or guaranteed ranking.

Reproducibility

Keep the original and prepared structures, topology, effective force-field versions, protonation choices, configuration, job metadata, logs, trajectory, and analysis scripts. Inspect temperature and energy behavior and assess sampling across time windows and independent runs. Automated job completion confirms execution, not scientific validity.

Input limitations to check explicitly

Discuss a system outside this scope before submitting it. Unsupported components can be dropped during preparation or produce an inappropriate model even when a job completes.