Essential Dynamics: A Tool for Efficient Trajectory Compression and Management

TitleEssential Dynamics: A Tool for Efficient Trajectory Compression and Management
Publication TypeJournal Article
Year of Publication2006
AuthorsMeyer, Tim, Ferrer-Costa Carles, Pérez Alberto, Rueda Manuel, Bidon-Chanal Axel, F. Luque Javier, Laughton Charles. A., and Orozco Modesto
JournalJournal of Chemical Theory and Computation
Volume2
Issue2
Pagination251 - 258
Date Published2006/03/01
ISBN Number1549-9618
Abstract

We present a simple method for compression and management of very large molecular dynamics trajectories. The approach is based on the projection of the Cartesian snapshots collected along the trajectory into an orthogonal space defined by the eigenvectors obtained by diagonalization of the covariance matrix. The transformation is mathematically exact when the number of eigenvectors equals 3N?6 (N being the number of atoms), and in practice very accurate even when the number of eigenvectors is much smaller, permitting a dramatic reduction in the size of trajectory files. In addition, we have examined the ability of the method, when combined with interpolation, to recover dense samplings (snapshots collected at a high frequency) from more sparse (lower frequency) data as a method for further data compression. Finally, we have investigated the possibility of using the approach when extrapolating the behavior of the system to times longer than the original simulation period. Overall our results suggest that the method is an attractive alternative to current approaches for including dynamic information in static structure files such as those deposited in the Protein Data Bank.We present a simple method for compression and management of very large molecular dynamics trajectories. The approach is based on the projection of the Cartesian snapshots collected along the trajectory into an orthogonal space defined by the eigenvectors obtained by diagonalization of the covariance matrix. The transformation is mathematically exact when the number of eigenvectors equals 3N?6 (N being the number of atoms), and in practice very accurate even when the number of eigenvectors is much smaller, permitting a dramatic reduction in the size of trajectory files. In addition, we have examined the ability of the method, when combined with interpolation, to recover dense samplings (snapshots collected at a high frequency) from more sparse (lower frequency) data as a method for further data compression. Finally, we have investigated the possibility of using the approach when extrapolating the behavior of the system to times longer than the original simulation period. Overall our results suggest that the method is an attractive alternative to current approaches for including dynamic information in static structure files such as those deposited in the Protein Data Bank.

URLhttps://dx.doi.org/10.1021/ct050285b
Short TitleJ. Chem. Theory Comput.
Highlight: 
Review: