author	Norwid Behrnd

aRMSD allows the manual the comparison of two constitutionally identical small molecule structures by the Kabsch algorithm¹. In addition to the original approach to account only for atomic positions, aRMSD provides adjustable weights, for instance to equally account for X-ray scattering factors (for instance Mo K(alpha) radition). The results of the superposition can be accessed as interactive renderings by vtk,² while statistics are visualized in diagrams generated by matplotlib.³ The development of the software was started prior to 2016 by Arne Wagner during his PhD thesis in the Himmel group (University of Heidelberg, Germany)⁴,⁵ and distributed on https://github.com/armsd/aRMSD under the MIT license scheme.

The current version 1.0.0 reorganizes the project for an easier installation and subsequent use with Python 3.10 (and above), agnostic to the operating system at hand. This is the reason why the primary input format of data to process should be .xyz, or (second best choice:) .pdb in the syntax written by OpenBabel.⁶ Contrasting to earlier releases, correct processing of .cif or output files by quantum mechanical programs (Gaussian, MOPAC, etc) can not be guaranteed.

What may aRMSD do for you – an appetizer

As shown in the figure below, aRMSD may be used to align the models (a) and reorder the atom order by the Hungarian algorithm (b). Subsequently, the superposition is optimized by minimization of the RMSD according to the Kabsch algorithm.

In a newly designed interactive representation (c), the differences between the two models are shown: atoms drawn with larger diameter indicate a larger relative contribution to the final RMSD determined for the complete model. Their color corresponds to the scale at the side about absolute positional differences (in Angstroms) of said atom in in reference and model in the optimized superposition.

In addition, aRMSD the program compares the corresponding bond lengths of model and reference indicates either by green or red color encoding if the one by the model is shorter, or lengthier than the corresponding bond in the reference. Not shown here, but aRMSD allows the interactive readout of the differences found for selected bond lengths, angles, and dihedral angles, too.

aRMSD equally provides a more classical (yet interactive) visualization of the optimized superposition (d). This "best fit" determined may be saved as a .xyz file, or as a set of 10 files about structure between model and reference.

In addition to an optional permanent log file (aRSMD_logfile.out) about setup and results of the superposition (e.g., final rmsd, cosine similarity, and GARD score⁷), the user may complement the scrutiny with diagrams. Generated by matplotlib, it is possible to pan and zoom regions of interest. The export includes formats like .png, .pdf, or .tikz.

Where stands aRMSD, relatively to other programs?

aRMSD allows a pair-wise comparison of small molecule structures with significant user-interaction. This offers you multiple levels to check and adjust the progress of the analysis. A more automatic analysis, potentially over batches of structures, is not foreseen; if interested, see for instance Jimmy Kromann's rmsd (equally implemented in Python).

Note, however, that an automated unsupervised scrutiny of model data may yield wrong results. One potential pitfall is how the model information is handled prior to the refinement of the structure alignment, where aRMSD uses the Hungarian algorithm. To quote Kildgaard:⁸

"The RMSD can be minimized by translating and rotating one set of coordinates (the other is held fixed) because the molecules are invariant under these operations. This will lead to the two molecules being superimposed but can also lead to a false RMSD value if the atoms are not ordered identically."

which was further demonstrated (and illustrated) for instance by Temeslo.⁹

Proposed installation and use

aRMSD preferably is used in a virtual environment of Python 3.11 (or above) to resolve its dependencies by pypi.org and file pyproject.toml. Depending on Python version and operating system, this support can add up to about 1GB permanent memory. With the .whl at disposition (release page of the GitHub repository, or built by uv build from the GitHub repository), the installation with pip provides armsd (all lower case) as new initial command to the CLI.

When running aRMSD from the CLI, ensure the terminal is tall enough (e.g., 40 rows instead of only 24; a width of 80 characters however is fine). Else you may miss some of its rolling interface, and commands at disposition. For details, see folder docs or the primer on armsd-primer.readthedocs.io.

Footnotes

Footnotes

1. Kabsch, W. A Solution for the Best Rotation to Relate Two Sets of Vectors. Acta Cryst. A 1976, 32 (5), 922–923. https://doi.org/10.1107/S0567739476001873. ↩

2. http://www.vtk.org ↩

3. https://matplotlib.org/ ↩

4. Wagner, A. Himmel, H.-J. aRMSD: A Comprehensive Tool for Structural Analysis. J. Chem. Inf. Model., 2017, 57, 428–438. https://doi.org/10.1021/acs.jcim.6b00516. ↩

5. Wagner, A. Synthese und Koordinationschemie guanidinatstabilisierter Diboranverbindungen. (Synthesis and Coordination Chemistry of Guanidinate-Stabilised Diboranes) PhD thesis (2015), University of Heidelberg (Germany). Written in German including an English summary. The pdf of this document may be found at the doi 10.11588/heidok.00019018. ↩

6. Open Babel, http://openbabel.org/wiki/Main_Page. For further details, see by O'Boyle, N. M.; Banck, M.; James, C. A.; Morley, C.; Vandermeersch, T.; Hutchison, G. R. Open Babel: An open chemical toolbox. J. Cheminf. 2011, 3:33. https://doi.org/10.1186/1758-2946-3-33. ↩

7. Baber, J. C.; Thompson, D. C.; Cross, J. B.; Humblet, C. GARD: A Generally Applicable Replacement for RMSD. J. Chem. Inf. Model. 2009, 49 (8), 1889–1900. https://doi.org/10.1021/ci9001074. ↩

8. Kildgaard, J. V.; Mikkelsen, K. V.; Bilde, M.; Elm, J. Hydration of Atmospheric Molecular Clusters: A New Method for Systematic Configurational Sampling. J. Phys. Chem. A 2018, 122, 5026–5036. https://doi.org/10.1021/acs.jpca.8b02758. ↩

9. Temeslo, B.; Mabey, J. M.; Kubota, T.; Appiah-Padi, N.; Shields, G. C. ArbAlign: A Tool for Optimal Alignment of Arbitrarily Ordered Isomers Using the Kuhn-Munkres Algorithm. J. Chem. Inf. Model. 2017, 57, 1045–1054. https://doi.org/10.1021/acs.jcim.6b00546. ↩