Practical System Preparation Tips for Molecular Dynamics Simulations

Mar 9, 2026·
Yassir Boulaamane
Yassir Boulaamane
· 3 min read

The trajectory is already biased before the first step. The box decides how much water you pay for. The termini decide whether a fragment carries charges it never had. The frame you pick afterwards decides whether you publish an outlier.

Three periodic boxes for the same soluteA cube is the simplest box and the least efficient for a globular protein. A truncated octahedron uses roughly 30 to 40 percent less solvent. A rhombic dodecahedron uses the smallest solvent volume. No percent is stated for the rhombic box.Same buffer, different water billCubeSimplestMost solventTruncatedoctahedronAbout 30–40%less solventRhombicdodecahedronSmallestvolume
Figure 1. The 30 to 40 percent saving is the one this note gives, and it is for a truncated octahedron around a roughly spherical solute, relative to a cube with the same minimum distance to the edge. The rhombic dodecahedron is described only as the most compact. A cube is still the right box when the solute is long along the Cartesian axes.

Which box

A cubic box is orthogonal, easy to look at, and easy to debug. For a globular protein it also packs badly: the corners fill with water that the solute never touches. That is a larger system and a higher cost for the same buffer.

A truncated octahedron is closer to a sphere. For spherical solutes it cuts the solvent count by roughly 30 to 40 percent and still wraps the solute in an even shell. It is the usual compromise.

A rhombic dodecahedron leaves the least empty solvent for a given buffer. The shape is less obvious. Current MD codes handle it. The saving matters once the solute is large. This note does not give a percent for that saving.

BoxAdvantageLimit
CubeSimple, and natural for an elongated soluteLeast efficient for a globular protein
Truncated octahedronBalance of simplicity and solvent countA less obvious geometry
Rhombic dodecahedronSmallest solvent volumeLeast intuitive shape

When to cap the termini

A full-length chain has a real N terminus and a real C terminus. A fragment cut out of a domain does not. The cut ends are charged as if they were the ends of the protein, and they are often buried in the parent structure. That charge invents salt bridges, pulls on the fold, and shows up later as “dynamics.”

The usual caps close the backbone. ACE is the acetyl cap on the N terminus. NME is the N-methylamide cap on the C terminus.

Cap a peptide, an isolated domain, or a stretch you extracted from a larger chain. Leave the termini alone on a full-length protein whose natural ends are solvent-exposed and part of the biology.

Pick a cluster centroid, not the farthest frameSeveral trajectories are aligned, clustered, and represented by centroids. The frame farthest from the start is an outlier, not the representative structure.The interesting frame is often the wrong onePool runsAlign a stable coreClusterRMSD, distances, pocketCentroidOne frame per stateFarthest from the start is an outlier
Figure 2. Pool the trajectories, align on a region that should not move, then cluster on RMSD, a few distances, or the pocket. The structure you keep is a centroid of a populated cluster. Distance from the first frame is not a population.

Which frame represents the ensemble

The frame that has wandered farthest from the start is easy to notice and a poor representative. It can be one excursion.

Pool the runs. Align on a stable reference region. Measure the variation you care about: RMSD, distances between key residues, or the pocket. Cluster. Take the centroids. Those are the conformations the ensemble actually spent time in.

Computational Chemistry Molecular Dynamics Drug Discovery
Yassir Boulaamane
Authors
Yassir Boulaamane
Postdoctoral Researcher