How to Visualize MD Trajectories in VMD and PyMOL: Complete Tutorial

VMD vs PyMOL: which should you use?

The practical answer is: use both. VMD for trajectory exploration and dynamic analysis; PyMOL for creating the figures that go in your paper. They serve different purposes and complement each other well. This tutorial covers both.

VMD is free and available at ks.uiuc.edu/Research/vmd. The open-source PyMOL version is installable via conda: conda install -c conda-forge pymol-open-source.

Before you visualize: prepare the trajectory

Raw GROMACS trajectories cannot be loaded directly into VMD or PyMOL without preparation. The periodic boundary conditions cause atoms to “jump” across box boundaries, splitting molecules across opposite faces of the simulation box. You must center and wrap the trajectory first.

If you haven’t already done this from the analysis tutorial, run:

# Step 1: fix PBC — make molecules whole across boundaries
gmx trjconv \
  -s md.tpr \
  -f md.xtc \
  -o md_whole.xtc \
  -pbc whole
# Select: System

# Step 2: center protein in box
gmx trjconv \
  -s md.tpr \
  -f md_whole.xtc \
  -o md_centered.xtc \
  -center \
  -pbc mol \
  -ur compact
# Center group: Protein
# Output group: System

You also need a GRO or PDB structure file to load alongside the trajectory — use your energy-minimized structure em.gro or convert it: gmx editconf -f em.gro -o structure.pdb. VMD and PyMOL both use this as the topology reference when reading the trajectory.

Visualizing in VMD

mol new structure.pdb
mol addfile md_centered.xtc type xtc waitfor all

Visualizing in PyMOL

# Load structure and trajectory
load structure.pdb, protein
load_traj md_centered.xtc, protein, interval=10

# interval=10 loads every 10th frame — adjust to manage file size
# Each loaded frame becomes a 'state' in PyMOL

# Clean base representation
hide everything
show cartoon, protein
show sticks, resname LIG
color gray80, protein
color cyan, resname LIG

# Show binding site residues as sticks
select binding_site, protein and (byres resname LIG around 5)
show sticks, binding_site
color wheat, binding_site

# Remove water from view (if still present)
hide everything, resname SOL HOH

# Set background and render settings
bg_color white
set ray_shadows, 0
set ambient, 0.4
set spec_reflect, 0.2

# Navigate to a specific frame (state number)
set state, 50   # frame 50

# Render high-resolution PNG
ray 2400, 1800  # width x height in pixels
png figure_frame50.png, dpi=300

Measuring distances over time

One of the most common things to measure during trajectory visualization is how the distance between two specific atoms or residues changes over time — for example, the distance between a ligand carboxylate oxygen and a protein arginine nitrogen, or the distance across a binding site.

In VMD

VMD can plot distances as a function of frame number using the Timeline plugin or the Distance label tool. The most reliable method for publication is the Tcl console:

# Measure distance between two atoms over all frames
# Replace 1234 and 5678 with actual atom indices from atom selection
set sel1 [atomselect top "resid 120 and name NH2"]
set sel2 [atomselect top "resname LIG and name O1"]

set outfile [open distances.dat w]
set nframes [molinfo top get numframes]

for {set i 0} {$i < $nframes} {incr i} {
    $sel1 frame $i
    $sel2 frame $i
    set d [measure bond [list [$sel1 get index] [$sel2 get index]]]
    puts $outfile "$i $d"
}
close $outfile
puts "Distances written to distances.dat"

In PyMOL

# Quick distance label between two atoms (updates with state changes)
distance d1, /protein//A/ARG`120/NH2, /protein//A/LIG/O1

# For distance over all states using Python in PyMOL
python
distances = []
for state in range(1, cmd.count_states('protein') + 1):
    d = cmd.get_distance('/protein//A/ARG`120/NH2',
                          '/protein//A/LIG/O1',
                          state=state)
    distances.append((state, d))

with open('pymol_distances.txt', 'w') as f:
    for s, d in distances:
        f.write(f'{s} {d:.3f}\n')
python end

Creating publication movies

In VMD

VMD’s Movie Maker plugin (Extensions → Visualization → Movie Maker) provides a GUI for rendering trajectory movies. Key settings for a publication-quality output:

• Renderer: Tachyon — ray-traced, significantly better quality than the default OpenGL renderer
• Format: MPEG-4 or a sequence of TIFF frames (assemble with ffmpeg for more control)
• Resolution: 1920×1080 minimum for journal submission; 3840×2160 for supplementary material

For fine-grained control, render frames individually from the Tcl console and assemble with ffmpeg:

ffmpeg \
  -framerate 30 \
  -i frame%04d.tga \
  -c:v libx264 \
  -pix_fmt yuv420p \
  -crf 18 \
  movie.mp4

In PyMOL

# Render all frames as a movie
set ray_trace_frames, 1    # ray trace each frame (slower but higher quality)
set cache_frames, 0         # don't cache — saves memory for long trajectories
mpng frames/frame_, ray=1   # saves frame_0001.png, frame_0002.png, etc.

Assemble the PNG frames with ffmpeg using the same command as above.

Common visualization mistakes

• Loading the raw trajectory without PBC correction
  The most common mistake. Atoms appear to teleport across the box, molecules split in half, and the protein looks like it’s falling apart. This is entirely a visualization artifact — the simulation is fine.
  Always run gmx trjconv with -pbc whole followed by -center -pbc mol before loading in VMD or PyMOL.
• Using the first frame as the publication figure
  The first frame is the starting crystal structure. It doesn’t represent the equilibrated simulation state. Reviewers and readers may reasonably ask how the figure frame was chosen.
  Choose a representative frame from the stable portion of the trajectory — typically after RMSD has plateaued. Describe how the representative frame was selected in the methods or figure caption.
• Showing water molecules in publication figures
  Displaying explicit water clutters the figure and makes it impossible to see the protein. There are almost no contexts in which showing water in a publication figure improves understanding.
  In VMD: use the selection not water. In PyMOL: hide everything, resname SOL HOH.
• Not smoothing the trajectory before visual inspection
  Raw MD trajectories are jerky — each frame shows thermal fluctuations. Watching the raw trajectory makes it hard to see genuine conformational events.
  In VMD, use Extensions → Analysis → Trajectory Smoother to apply a running average. In PyMOL, load every Nth frame with the interval parameter in load_traj. For movies, smoothing 5–10 frames produces much more interpretable visualization.
• Measuring distances in the unprocessed trajectory
  A distance measurement between a protein residue and a ligand on opposite sides of a periodic boundary will give a large, meaningless value (the box dimension rather than the true distance).
  Always measure distances in the centered, PBC-corrected trajectory. For critical measurements, also verify with gmx distance on the processed trajectory — it handles PBC correctly regardless.