Predictive compression of molecular dynamics trajectories
Molecular dynamics simulations help to understand the complex behavior of molecules. The output of such a simulation describes the trajectories of individual atoms as snapshots of atom positions in time. Many compression schemes were developed to reduce the amount of data needed for storing long tra...
Saved in:
Published in | Journal of molecular graphics & modelling Vol. 96; p. 107531 |
---|---|
Main Authors | , , |
Format | Journal Article |
Language | English |
Published |
United States
Elsevier Inc
01.05.2020
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | Molecular dynamics simulations help to understand the complex behavior of molecules. The output of such a simulation describes the trajectories of individual atoms as snapshots of atom positions in time. Many compression schemes were developed to reduce the amount of data needed for storing long trajectories. This is achieved by limiting the precision of coordinates, encoding differences instead of absolute values, dimensionality reduction by principal component analysis, or by using polynomials approximating vertex trajectories. However, compression schemes using actual bonds between atoms have not been utilized to their full potential. Therefore, we developed a lossy compression method that captures the local, mostly rotational movement of atoms with respect to their bonded neighbors and predicts their positions in each frame. This allows full control over the data distortion. In our experiments, the method achieves data rates which are substantially better than the rates achieved by competing methods at the same error level.
•Content aware lossy predictive compression method for dynamic macromolecules proposed.•It decouples the movement occurring due to changing dihedral angles.•The method provides full control over the maximal error in atom coordinates.•In our experiments, it outperforms other state-of-the-art methods. |
---|---|
Bibliography: | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 23 |
ISSN: | 1093-3263 1873-4243 |
DOI: | 10.1016/j.jmgm.2020.107531 |