Skip to main content
The four core analyses that handle 80% of molecular dynamics workflows. Master these patterns and your agent can tackle most trajectory analysis tasks.

​
Radius of Gyration (Rg)

Question answered: How compact is this structure? 
from warp_md import System, RgPlan

system = System.from_pdb("protein.pdb")
traj = system.open_trajectory("trajectory.xtc")

plan = RgPlan(
    selection=system.select("name CA"),
    mass_weighted=False  # or True for mass-weighted Rg
)

rg = plan.run(traj, system, device="auto")

print(f"Shape: {rg.shape}")        # (n_frames,)
print(f"Mean Rg: {rg.mean():.2f} Å")
Output: 1D array with one Rg value per frame. Simple, clean, and GPU-accelerated.

​
Root Mean Square Deviation (RMSD)

Question answered: How much has the structure changed from the reference? 
from warp_md import RmsdPlan

plan = RmsdPlan(
    selection=system.select("backbone"),
    reference="topology",  # compare against initial structure
    align=True             # superimpose before measuring
)

rmsd = plan.run(traj, system, device="auto")

print(f"Final RMSD: {rmsd[-1]:.2f} Å")
ReferenceWhat It Compares Against
"topology"Initial PDB/GRO coordinates
"frame0"First trajectory frame

​
Mean Square Displacement (MSD)

Question answered: How far are molecules diffusing? 
from warp_md import MsdPlan

plan = MsdPlan(
    selection=system.select("resname SOL and name OW"),
    group_by="resid",  # treat each water molecule as a unit
)

time, msd = plan.run(traj, system, device="auto")

print(f"Time points: {len(time)}")
print(f"MSD shape: {msd.shape}")

​
Advanced MSD Options

plan = MsdPlan(
    selection=system.select("resname SOL"),
    group_by="resid",
    axis=[0.0, 0.0, 1.0],              # project onto z-axis
    frame_decimation=(100, 100),       # start at frame 100, stride 100
    lag_mode="multi_tau",              # bounded memory for long trajectories
)
ModeDescriptionWhen to Use
"auto"FFT if fits memory, else multi-tauDefault (let warp-md decide)
"multi_tau"Log-spaced lags, bounded memoryLong trajectories
"ring"Exact up to max_lagShort-time validation
"fft"Exact all lagsShort trajectories

​
Radial Distribution Function (RDF)

Question answered: What’s the local structure around each atom? 
from warp_md import RdfPlan

plan = RdfPlan(
    sel_a=system.select("resname SOL and name OW"),
    sel_b=system.select("resname SOL and name OW"),
    bins=200,
    r_max=10.0,
    pbc="orthorhombic"
)

r, g, counts = plan.run(traj, system)

print(f"First peak at r = {r[g.argmax()]:.2f} Å")
Output:
  • r: bin centers (Å)
  • g: g(r) values (normalized radial distribution)
  • counts: raw pair counts

​
Complete Example: All Four at Once

from warp_md import System, Trajectory
from warp_md import RgPlan, RmsdPlan, MsdPlan, RdfPlan

# Load system and trajectory
system = System.from_pdb("solvated_protein.pdb")
traj = Trajectory.open_xtc("production.xtc", system)

# Define selections
protein = system.select("protein")
backbone = system.select("backbone")
water_o = system.select("resname SOL and name OW")

# Run all four analyses
rg = RgPlan(protein).run(traj, system)
rmsd = RmsdPlan(backbone, reference="topology", align=True).run(traj, system)
time, msd = MsdPlan(water_o, group_by="resid").run(traj, system)
r, g, _ = RdfPlan(water_o, water_o, bins=200, r_max=10.0).run(traj, system)

# Results
print(f"Rg: {rg.mean():.2f} ± {rg.std():.2f} Å")
print(f"RMSD: {rmsd.mean():.2f} ± {rmsd.std():.2f} Å")
print(f"RDF first peak: {r[g.argmax()]:.2f} Å")
Four comprehensive analyses in 15 lines. GPU-accelerated when available.

​
JSON Config for Batch Analysis

{
  "version": "warp-md.agent.v1",
  "run_id": "protein-analysis",
  "system": {"path": "solvated_protein.pdb"},
  "trajectory": {"path": "production.xtc"},
  "device": "auto",
  "output_dir": "outputs",
  "analyses": [
    {
      "name": "rg",
      "selection": "protein",
      "mass_weighted": false
    },
    {
      "name": "rmsd",
      "selection": "backbone",
      "reference": "topology",
      "align": true
    },
    {
      "name": "msd",
      "selection": "resname SOL and name OW",
      "group_by": "resid",
      "lag_mode": "multi_tau"
    },
    {
      "name": "rdf",
      "sel_a": "resname SOL and name OW",
      "sel_b": "resname SOL and name OW",
      "bins": 200,
      "r_max": 10.0
    }
  ]
}

warp-md run config.json --stream ndjson

​
Performance Tips

Use GPU

Set device="auto" to use CUDA when available. Typical speedups: 8-15x for Rg/RDF/MSD.

Chunk Frames

For large trajectories, adjust chunk_frames to optimize memory:
plan.run(traj, system, device="cuda", chunk_frames=1000)

Multi-tau for MSD

Use lag_mode="multi_tau" for long trajectories to avoid memory issues.

Batch Analyses

Load trajectory once, run multiple analyses. Use config runner for complex workflows.

​
See Also

Build docs developers (and LLMs) love

Get started for freeTalk to us