BitClust: Fast Geometrical Clustering of Long Molecular Dynamics Simulations.
J Chem Inf Model
; 60(2): 444-448, 2020 02 24.
Article
em En
| MEDLINE
| ID: mdl-31651166
ABSTRACT
The growing computational capacity allows the investigation of large biomolecular systems by increasingly extensive molecular dynamics simulations. The resulting huge trajectories demand efficient partition methods to discern relevant structural dissimilarity. Clustering algorithms are available to address this task, but their implementations still need to be improved to gain in computational speed and to reduce the consumption of random access memory. We propose the BitClust code which, based on a combination of Python and C programming languages, performs fast structural clustering of long molecular trajectories. BitClust takes advantage of bitwise operations applied to a bit-encoded pairwise similarity matrix. Our approach allowed us to process a half-million frame trajectory in 6 h using less than 35 GB, a task that is not affordable with any of the similar alternatives.
Texto completo:
1
Coleções:
01-internacional
Base de dados:
MEDLINE
Assunto principal:
Algoritmos
/
Simulação de Dinâmica Molecular
Idioma:
En
Ano de publicação:
2020
Tipo de documento:
Article