Rowan Docs

Protein Molecular Dynamics

Python API · stjames models · API example

How it works

Simulate an apo protein, protein–ligand or protein–protein complex in explicit water with OpenMM. Compare replicate trajectories to assess structural stability, flexibility, and binder motion over the simulated timescale.

Settings

  • System: provide a prepared protein, retain required cofactors and structural ions, and supply SMILES for small molecules requiring parameterization. Optionally designate binder chains or small molecules; other small molecules remain receptor cofactors. See tricky cases in protein preparation.
  • Production: “Num trajectories” and “Simulation time (ns)” default to four 10 ns trajectories in the web form.
  • Sampling: “Frame save interval (ps)” defaults to 10; “Analysis interval (ps)” defaults to 100 for receptor solvent-accessible surface area (SASA), polar SASA, and binder MM/GBSA. Leave the analysis interval blank to disable these analyses.
  • Temperature: “Temperature (K)” defaults to 300.
  • Downloads: “Save solvent?” is off by default. With a binder, “Max solvent to save” retains the nearest waters at each frame; blank retains all solvent when saving solvent is enabled.
  • Restraints: “Protein restraint cutoff (Å)” defaults to 7 in the web form. Receptor Cα atoms farther from the binder are restrained; blank or no binder disables restraints.
  • Clustering: “Trajectory clustering” defaults to “None.” “K-means” uses “Num clusters” (initially 10); “Greedy” uses “Cutoff (Å)” (initially 2).
  • Advanced settings: defaults are AMBER ff14SB, OpenFF Sage 2.3.0, TIP3P water, 0.5 ns equilibration, 1 atm pressure, a 1 ps Langevin timescale, a 4 fs timestep, 3 Da hydrogen masses, 0 M added salt, and 8 Å water buffer and nonbonded cutoff. Hydrogen constraints are enabled.

Notes

The system is minimized, heated, and equilibrated at constant pressure and temperature before production. Replicas begin from the same equilibrated structure and velocities. Compare replicas; a stable short trajectory does not establish converged sampling or binding affinity. Restraints suppress motion outside the binder region, so consider their effect when interpreting flexibility.

API defaults differ: one trajectory, no SASA/MM/GBSA analysis, and no protein restraints unless requested. The default 4 fs timestep relies on hydrogen constraints and mass repartitioning; changing these requires an appropriate timestep.

Protein RMSD measures all protein Cα atoms, including binder chains, relative to the first saved frame after alignment. RMSF measures per-Cα fluctuations; radius of gyration measures receptor polymer compactness. Binder RMSD follows receptor alignment and is available only for one small molecule (heavy atoms) or one binder chain (backbone atoms). Multiple binder components have no combined RMSD.

MM/GBSA reports interaction energies in kcal/mol using GBn2 implicit solvent, retaining receptor cofactors. It omits binding entropy and is not a binding free energy. Hydration-site analysis pools binder-interface waters across replicas, including when downloaded trajectories omit solvent; see hydration-site case studies.

Downloads include the minimized structure and trajectories. Average structures may not be physically sampled conformations; the medoid is an actual representative frame. Optional clustering provides frame assignments and representative frame indices.

Benchmarks and validation

As a representative benchmark, a single-replica run on a dasatinib·ABL kinase system (PDB 2GQG) with 55,517 atoms solvated, parameterized, minimized, equilibrated, and completed a 10-ns production trajectory in 18 minutes total, reaching speeds of >1 µs/day on our standard L40S GPU hardware.

Further reading