A Python package for the characterisation of local chemical environment dynamics for Molecular Dynamics trajectories of proteins and other biomolecules.

SPEADI provides the user tools with which to characterise the local chemical environment using Molecular Dynamics data. At the moment, it implements two variations of the pair radial distribution function (RDF): time-resolved RDFs (TRRDFs) and van Hove dynamic correlation functions (VHFs).

For installation into the default python environment, run the following pip command in a terminal:

pip install git+https://github.com/FZJ-JSC/speadi.git

Or, to install just into the current user's local environment, add the --user option:

pip install --user git+https://github.com/FZJ-JSC/speadi.git

Normally, Radial Distribution Functions used in atomistic simulations are averaged over entire trajectories. SPEADI averages over user-defined windows of time. This gives a separate RDF between group a and b for each window in the trajectory.

This package provides a method to calculate Time-resolved Radial Distribution Functions using atomistic simulation trajectory data. This is implemented efficiently using Numpy arrays and the Numba package when available. Trajectory data may be anything that the package MDTraj can handle, or preferably a string pointing to the location of such data.

The trajectory data used must be sampled to a sufficient degree. The sampling frequency of the input data naturally determines the time-resolving capability of the time-resolved RDF. When using both small groups for reference particles and selection particles, the sampling frequency must be high enough to provide an ensemble of distances in each time slice that provides a satisfactory signal-to-noise ratio.

In RDFs of particular ions around selected single atoms in all-atom simulations of biomolecules, experience suggests a sampling frequency of below 1 picosecond and window lengths of 1-10 picoseconds to follow changes in coordination shells of atoms.

This package also provides a method to calculate the van Hove dynamic correlation function using atomistic simulation trajectory data. This is also implemented efficiently using Numpy arrays and the Numba package when available. Trajectory data may be anything that the package MDTraj can handle, or preferably a string pointing to the location of such data.

As with TRRDFs, the input trajectory data must be sampled above a certain frequency. Window lengths of 1-2 picoseconds are enough to follow the loss in structure in all-atom simulations of water, with each window containing anything above 10 samples.

VHFs of ions around single atoms in do not require sample frequencies above those consistent with the time-scale of the movement of ions, yet they do require a larger number of windows to average over. This number of windows can either be supplied by using a trajectory or trajectory slice of preferably at least 10 ns in length, or alternatively by using a sliding window over a trajectory slice of at least 1 ns in length.

Installation is provided easily through pip. It can be installed either directly as a package, or as an editable source.

As a default, SPEADI doesn't install JAX or Numba, but uses these if detected in the same Python environment that SPEADI is installed into.

To install JAX and jaxlib along with SPEADI, simply add the jax extra to pip:

pip install 'git+https://github.com/FZJ-JSC/speadi.git#egg=SPEADI[jax]'

Note that by default, installing jax using pip (through pypi) only enables CPU acceleration. To enable GPU or TPU acceleration, please see https://github.com/google/jax for details on how to obtain a JAX installation for the specific CuDNN version in your environment.

To install Numba along with SPEADI, simply add the numba extra to pip:

pip install 'git+https://github.com/FZJ-JSC/speadi.git#egg=SPEADI[numba]'

Or, to install both jax and numba alongside SPEADI, add the all extra to pip:

pip install 'git+https://github.com/FZJ-JSC/speadi.git#egg=SPEADI[all]'

The --user pip option may be added to all of these commands to install just for the current user.

Open up a terminal. Navigate to the location you want to clone this repository. Then, run the following to clone the entire repository:

git clone https://github.com/FZJ-JSC/speadi

Then, install locally using pip by adding the -e option:

pip install -e speadi

To calculate the time-resolved RDF for every single protein heavy atom with each ion species in solvent, you first need to specify the trajectory and topology to be used:

topology = './topology.gro'
trajectory = './trajectory.xtc'

Next, load the topology in MDTraj and subset into useful groups:

import mdtraj as md

top = md.load_topology(topology)
na = top.select('name NA')
cl = top.select('name CL')
protein_by_atom = [top.select(f'index {ix}') for ix in top.select('protein and not type H')]

Now you can load SPEADI to obtain RDFs:

import speadi as sp

To make an RDF for each heavy protein atom

r, g_rt = sp.trrdf(trajectory, protein_by_atom, [na, cl], top=top, n_windows=1000, window_size=500,\
              skip=0, pbc='general', stride=1, nbins=400)

To repeat the analysis, but obtain integral of $g(r)$ instead, simply replace trrdf with int_trrdf instead.

r, n_rt = sp.int_trrdf(trajectory, protein_by_atom, [na, cl], top=top, n_windows=1000, window_size=500,\
              skip=0, pbc='general', stride=1, nbins=400)

First, load a topology as above using MDTraj, then define the reference and target groups:

import mdtraj as md

top = md.load_topology(topology)
water_H = top.select('name HW2')
target_atom = top.select('resid 129 and name OG')

In this example, we're looking at the stability of the water structure surrounding the side-chain terminal oxygen in a serine residue.

Next, calculate the VHF for this site over the whole trajectory:

r, G_s, G_d = sp.vanhove(trajectory, target_atom, [water_H], 
                        top=top, n_windows=1000, window_size=500,
                        skip=0, pbc='general', stride=1, nbins=400)

As the reference and target particles are non-identical, $G_s$ is empty. The distinct part, $G_d$, gives us the time-dependent dynamic correlation between the two types of particles.

If you're using SPEADI for academic work, please cite both the following original paper published in Biology, as well as the Zenodo DOI for the version that you are using.

de Bruyn, E., Dorn, A. E., Zimmermann, O., & Rossetti, G. (2023). SPEADI: Accelerated Analysis of IDP-Ion Interactions from MD-trajectories. Biology, 12(4), 581. https://dx.doi.org/10.3390/biology12040581

Linked online version: https://www.mdpi.com/2079-7737/12/4/581

Zenodo DOI for the current SPEADI version: https://doi.org/10.5281/zenodo.7436713

In due time, a list of published papers that contain analysis performed using SPEADI will appear here.

We gratefully acknowledge the following institutions for their support in the development of SPEADI and for granting compute time to develop and test SPEADI.

• Gauss Centre for Supercomputing e.V. (www.gauss-centre.eu) and the John von Neumann Institute for Computing (NIC)

on the GCS Supercomputer JUWELS at Jülich Supercomputing Centre (JSC)

• HDS-LEE Helmholtz Graduate School

• Emile de Bruyn ([email protected])

SPEADI Copyright (C) 2022 Forschungszentrum Jülich GmbH, Jülich Supercomputing Centre and the Authors

This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2.1 of the License, or (at your option) any later version.

This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.