MolAlignLib is a Fortran and Python library based on random rotations and quasi-local RMSD minimizations to align rigid molecules and clusters. The details of the method can be found in publication [1].

Before installing

If you can't install stuff on your computer then try MolAlignLib on Binder.

• To build the native executable you will need a Fortran 2008 compiler and LAPACK.

• To build the python module you will need GFortran 4.8 or higher, LAPACK, Python 3.6 or higher, NumPy and ASE.

The easiest way to install the required packages is with your distro's package manager:

RHEL, Centos, Fedora, etc.

yum install git python3 python3-pip gcc-gfortran lapack-devel

Debian, Ubuntu, Mint, etc.

apt install git python3 python3-pip gfortran liblapack-dev

Build the native executable

Clone the repository:

git clone https://github.com/qcuaeh/molalignlib.git

then enter the cloned directory, edit the build.env file to suit your system and run:

./build.sh

It will create the molalign executable inside the build directory.

or build and install the python module

Simply run:

pip3 install git+https://github.com/qcuaeh/molalignlib.git

It will install NumPy, ASE and MolAlignLib in your site packages and the molalign script in your path.

Program options

Options supported by the native executable and the python script

-sort  Reorder atoms to minimize the RMSD.
-fast  Prune assignments that surpass the displacement tolerance.
-tol TOL  Set the displacement tolerance to TOL (defaults to 0.35 Å).
-out NAME  Set the output file name to NAME (defaults to aligned.xyz).
-count N  Set the count threshold to N (defaults to 10).
-trials N  Set the maximum number of trials to N.
-rec N  Record up to N assignments (defaults to 1).
-stats  Print detailed stats of the calculation.
-test  Use a fixed random seed for testing.
-mirror  Reflect aligned coordinates.
-mass  Use mass weighted RMSD.

Options only supported by the native executable

-live  Print stats in real time (if stats are enabled).
-stdin FMT  Read coordinates from standard input with format FMT.
-stdout FMT  Write coordinates to standard output with format FMT.

Basic usage

The syntax of the command is:

molalign [option[s]] file1 file2

The coordinates in file2 will be aligned to the coordinates in file1. If there is more than one set of coordinates in a file, only the first one will be read. The native executable only reads xyz and mol2 files while the python script reads all the file formats supported by ASE.

• Running the command without options will align the atoms without reordering.

• Running the command with the -sort option will remap the atoms to minimize the RMSD and the aligned coordinates will be written in the optimal mapping order.

Advanced usage

When reordering is performed the computation can take a lot of time to complete but can be considerably speeded up with the -fast option which enables pruning of any assignment that surpass the displacement tolerance. However, if the atom displacements are larger than this tolerance, the assignment will fail, or, if they are very close, the assignment can be suboptimal. In such cases the displacement tolerance should be increased with the -tol option.

The count threshold is used to decide if the procedure is converged. A threshold of 10 counts is used by default, which works fine for almost all cases, but you can change it with the -count option. To avoid too long computations you can set a maximum number of random trials with the -trials option, the computation will be aborted if it is reach before the count threshold.

The algorithm always explores multiple possible assignments, but only the best one is recorded by default. To record additional assignments use the -rec option. To print the stats of the computation use the -trials option, but notice that they will be different on each repeated run due to the use of randomized random seeds.

Command line examples

For small atom displacements the default tolerance is enough:

./build/molalign examples/Co138_0.xyz examples/Co138_1.xyz -sort -fast

Optimized RMSD = 0.0506

but for atom displacements larger than the tolerance the assignment will fail:

./build/molalign examples/Co138_0.xyz examples/Co138_3.xyz -sort -fast

Error: Assignment failed

Increasing the tolerance will fix the problem but the calculation will slow down:

./build/molalign examples/Co138_0.xyz examples/Co138_3.xyz -sort -fast -tol 0.7

Optimized RMSD = 0.1973

Printing multiple alignments and stats can be useful to identify rotational symmetric clusters:

./build/molalign examples/Co138_0.xyz examples/Co138_1.xyz -sort -fast -rec 5 -stats

 Map    Count    Steps     Angle       RMSD
-------------------------------------------
   1       10     12.1      60.5     0.0506
   2        9     12.3      66.4     0.0506
   3       15     10.1      50.2     0.0506
   4        1      9.0      41.8     0.6652
   5        1      6.0       4.1     0.6716
-------------------------------------------
Random trials = 66
Minimization steps = 595
Visited local minima > 5

There are three different assignments with the same RMSD indicating that the cluster has three fold symmetry. Notice that the three optimal symmetric assignments are visited multiple times while the suboptimal ones are visited only once.

Python examples

Minimize the RMSD between two Cobalt clusters:

from ase.io import read
from molalignlib import assign_atoms
mol0 = read('Co138_0.xyz')
mol1 = read('Co138_1.xyz')
# Find optimal assignment
assignment = assign_atoms(mol0, mol1, fast=True)
# Reorder mol1 with the optimal assignment
mol1 = mol1[assignment.order]
# Align mol1 to mol0 (returns RMSD)
mol1.align_to(mol0)

MolAlignLib uses ASE to read and write atomic coordinates.

References

[1] J. M. Vasquez-Perez, L. A. Zarate-Hernandez, C. Z. Gomez-Castro, U. A. Nolasco-Hernandez. A Practical Algorithm to Solve the Near-Congruence Problem for Rigid Molecules and Clusters, Journal of Chemical Information and Modeling (2023), DOI: https://doi.org/10.1021/acs.jcim.2c01187