MD Trajectory Analysis

What is MD trajectory analysis?

MD trajectory analysis turns a simulation trajectory into measurements that can actually be interpreted: structural drift, residue flexibility, compactness, contact persistence, secondary-structure content, collective motions, and solvent-facing behavior. The raw frames coming out of GROMACS, AMBER, CHARMM, NAMD, or OpenMM do not answer those questions on their own.

This ProteinIQ tool runs a workflow pinned to MDAnalysis 2.9.0. It is not a single command-line program. It is a curated analysis surface built from MDAnalysis modules and a small amount of integration logic that standardizes inputs, runs multiple observables in one job, and returns structured outputs for charting and download.

The practical value is in combining observables that answer different failure modes:

• RMSD: Detects whether the selected atoms stay close to a chosen reference structure.
• RMSF: Shows which residues are mobile after aggregating atom-level fluctuations by residue.
• Radius of gyration: Tracks compaction or expansion over time.
• Distance tracking: Follows a specific geometric separation, often between domains, termini, ligands, or catalytic motifs.
• Hydrogen bonds and salt bridges: Measures persistence of stabilizing interactions.
• PCA and clustering: Describes dominant motions and recurring conformational states.
• DSSP, Ramachandran, chi angles, helix analysis: Connects motion to secondary structure and torsional states.
• SASA, RDF, density, water dynamics: Relates structural changes to solvent exposure and local environment.

Applications

• Fold stability checks: Compare early and late frames after simulations initialized from AlphaFold 2, Boltz-2, or ESMFold.
• Mutation analysis: Test whether a variant shifts flexibility, compaction, hydrogen-bond persistence, or helix geometry relative to the wild type.
• Binding-site monitoring: Track contact counts, hydrogen bonds, or a distance restraint around a ligand or interface during refinement.
• Ensemble characterization: Use PCA and clustering to summarize whether the trajectory samples one basin or several distinct states.
• Solvent-facing transitions: Use SASA, RDF, density, and water dynamics to examine collapse, exposure, or hydration changes.

How to use MD trajectory analysis online

Upload a topology file and a matching trajectory file, choose one or more analyses, and run the job to receive tabular and chart-ready outputs for each selected observable. The result page returns structural, interaction, conformational, environmental, and secondary-structure summaries with downloadable data for follow-up analysis.

Inputs

Common pairings include:

• .gro + .xtc for GROMACS
• .tpr + .xtc when topology metadata from GROMACS should be preserved
• .prmtop + .nc, .ncdf, or .mdcrd for AMBER
• .psf + .dcd for CHARMM or NAMD
• .pdb, .crd, or .gro with coordinate trajectories such as .xyz when atom ordering matches

How does MD trajectory analysis work?

The tool loads the uploaded files into an MDAnalysis Universe, applies the selected atom selection, resolves the reference mode, optionally aligns sampled frames to that reference, and then dispatches the requested observables.

Three details matter because they change interpretation:

• Reference mode: first, last, and average are not cosmetic options. average computes an average reference over the sampled frames instead of aliasing to the first frame.
• Frame sampling: step subsamples the trajectory globally for the audited analysis paths, which reduces runtime and directly changes temporal resolution.
• Alignment: When enabled, sampled frames are superimposed onto the chosen reference before RMSD, RMSF, PCA, clustering, and cross-correlation calculations. This removes rigid-body translation and rotation and makes those outputs more about internal motion than bulk drift.

Some analyses rely on additional assumptions from the underlying MDAnalysis modules or the tool:

• Hydrogen bonds: The tool uses a donor-acceptor cutoff of 3.0 Å and angle cutoff of 150°, matching the audited MDAnalysis default. Explicit hydrogens are still required for meaningful hydrogen-bond counts.
• H-bond lifetime: Lifetime analysis uses the same MDAnalysis hydrogen-bond machinery and respects the selected atom group. A protein and name CA selection is useful for fold tracking but does not contain hydrogens, so lifetime analysis will usually return a warning unless a chemically complete selection is used.
• Contacts: Contact analysis uses MDAnalysis.analysis.contacts.Contacts and reports the source native-contact fraction Q. The displayed contact count is a compatibility summary derived from that fraction and the first-frame native contact set.
• MSD: Mean square displacement uses MDAnalysis.analysis.msd.EinsteinMSD with msd_type="xyz". Physical diffusion estimates require trajectories in an unwrapped coordinate convention.
• RMSF: MDAnalysis computes atom-level fluctuations. The tool then aggregates them by residue so the output is residue-oriented even when the selection contains more than one atom per residue.
• DSSP: Secondary-structure assignment uses MDAnalysis.analysis.dssp when available. If that path fails, the tool falls back to a simpler internal estimate rather than aborting the whole job.
• Chi angles and helix geometry: Side-chain χ1/χ2 summaries use MDAnalysis Janin analysis with circular angle statistics. Helix geometry uses MDAnalysis HELANAL on alpha-carbon selections and requires consecutive helix segments of at least 9 residues.
• Water dynamics: This mode is available in the current tool, but it depends on water naming conventions and native MDAnalysis functionality that is being deprecated on the path to MDAnalysis 3.0.

Understanding the settings

Core settings

Advanced settings

Selection examples

Understanding the results

The result page groups outputs into tabs. Not every selected analysis appears in every tab, so the most useful way to read the output is by question rather than by file format.

Structural outputs

Interaction outputs

Conformational outputs

Secondary structure and environment outputs

Interpreting results

RMSD and RMSF

RMSD is easiest to misread when alignment and atom selection are not considered together.

• Low, stable RMSD after alignment usually indicates internal structural stability of the selected atoms.
• Low RMSD without alignment disabled can hide large rigid-body motion if the entire selection rotates or translates together.
• High RMSF in loops or termini is often expected.
• High RMSF inside a buried core helix or beta strand usually deserves inspection, especially when it appears alongside SASA growth or DSSP loss.

Distance, contacts, and hydrogen bonds

Distance tracking is strongest when paired with contacts or hydrogen bonds.

• Distance increases while contact count drops often indicates real separation rather than internal breathing.
• Stable distance with fluctuating contacts can mean the interface remains nearby but repacks locally.
• Hydrogen bonds near zero with an explicit warning usually means the uploaded structure lacks hydrogens, not necessarily that the system has no hydrogen bonding in reality.

PCA and clustering

PCA answers whether motion is dominated by a few directions. Clustering answers whether those motions produce recurring states.

• High PC1 variance with one dominant cluster suggests one major breathing or hinge motion inside one broad state.
• Several populated clusters with separated PC projections suggests discrete conformations rather than one continuous fluctuation cloud.
• Poorly separated clusters often mean the trajectory is either short, noisy, or not naturally state-separated under the chosen atom selection.

DSSP, SASA, and compactness

These metrics become more informative when read together.

• Rg decreases while SASA decreases usually indicates compaction.
• DSSP helix or sheet content falls while RMSF rises suggests local unfolding or disordering.
• SASA rises without major Rg change often reflects side-chain or surface rearrangement rather than global unfolding.

When to use MD trajectory analysis vs alternatives

MD trajectory analysis on ProteinIQ is a post-simulation interpretation tool. It is most useful after a trajectory already exists.

The main decision point is whether the question is dynamic or static:

• For one structure or one pairwise comparison, a simpler structure-analysis tool is usually enough.
• For questions about persistence, transitions, convergence, flexibility, or hydration over time, trajectory analysis is the correct layer.

Caveats that matter in practice

• Sampling frequency still limits interpretation: A frame saved every 10 ns cannot recover fast side-chain switching, no matter how many downstream analyses are run.
• Selection choice changes the science: protein and name CA is useful for global fold tracking, but it can miss side-chain rearrangements and interaction chemistry.
• Hydrogen-dependent analyses depend on protonation and topology quality: Missing hydrogens or inconsistent naming can suppress hydrogen-bond signals.
• MSD assumes unwrapped coordinates: Wrapped trajectories can make diffusion estimates physically misleading even when the calculation succeeds.
• HELANAL requires alpha-carbon helix segments: Short selections or non-consecutive residue groups may return a warning instead of helix geometry.
• Subsampling trades speed for detail: Larger step values reduce runtime and noise, but short-lived events may disappear.
• Water dynamics is not future-proof: The current tool still supports it, but the underlying MDAnalysis path is being deprecated on the route to version 3.0.