Path A - Module 47: Trajectory Management (Slicing)#

Real simulations produce gigabytes of data spread across multiple files. Managing these structures efficiently without losing track of time is a developer’s challenge.

In this module, you will learn to slice, dice, and concatenate your Alzheimer’s trajectories using MolSysMT.

import molsysmt as msm
from molsysmt import systems

# Let's use our Villin trajectory as a proxy for a long simulation
villin_traj = systems['chicken villin HP35']['traj_chicken_villin_HP35_solvated.dcd']
villin_topo = systems['chicken villin HP35']['traj_chicken_villin_HP35_solvated.h5msm']
molsys = [villin_topo, villin_traj]

1. Structure Slicing#

You don’t always need every structure. You can use the structure_indices argument in almost every function to focus on a specific time window.

# Get the number of structures
n_structures = msm.get(molsys, element='system', n_structures=True)
print(f"Total structures: {n_structures}")

# Extract only the first 10 structures and the last structure
indices = list(range(10)) + [n_structures - 1]
short_traj = msm.extract(molsys, structure_indices=indices)

print(f"Structures in short_traj: {msm.get(short_traj, element='system', n_structures=True)}")

2. Concatenating Structures#

If you have two separate simulation runs, you can join them into a single virtual object using concatenate_structures(). Coordinate-only trajectory parts do not need to duplicate the topology, but they must use the same atom count and ordering.

# Split our short_traj into two parts
part1 = msm.extract(short_traj, structure_indices=[0, 1, 2])
part2 = msm.extract(short_traj, structure_indices=[3, 4, 5])

# Concatenate them back
merged_traj = msm.concatenate_structures([part1, part2])

print(f"Structures in merged_traj: {msm.get(merged_traj, element='system', n_structures=True)}")

3. Sub-sampling (Striding)#

A common task is to take every Nth structure to reduce the data size for visualization or quick analysis.

# Take every 10th structure of the original system
sub_sampled = msm.extract(molsys, structure_indices=range(0, n_structures, 10))

print(f"Original: {n_structures} structures -> Sub-sampled: {msm.get(sub_sampled, element='system', n_structures=True)} structures")

Tier-1 XYZ, DCD, and XTC conversion routes have exhaustive coordinate-trajectory reports. Atom selections are materialized in canonical increasing order, while structure_indices keeps the requested order; target formats that cannot carry a value report the omission explicitly.

🏆 Path A Challenge: The Data Curator#

  1. Create a trajectory that contains only the even structures (0, 2, 4…) of the Villin simulation.

  2. Append the last 5 structures of the simulation to the end of your new trajectory.

  3. Use msm.get(..., structure_id=True) to see if the structure IDs are preserved or if they are reindexed.

Managing data is the boring but necessary part of pipeline development. In Module 48, we will see the “Heavy Artillery”: how to process files that are larger than your RAM using Iterators and the ChunkedExecutor.