![Show Menu](styles/mobile-menu.png)
![Page Background](./../common/page-substrates/page0048.jpg)
Conformational Ensembles from Experimental Data
and Computer Simulations
Poster Abstracts
44
11-POS
Board 11
Density of State Estimates for Conformational Ensembles Using an Improved Wang-
Landau Algorithm
Augustin Chevallier
, Frederic Cazals.
INRIA, Sophia Antipolis, France.
Generalized ensemble methods have proven effective to explore the energy landscape of
complex systems. Among the available options, the Wang-Landau (WL) algorithm implements a
random walk in energy space, from which the density of states is obtained.
However, even though several improvements have been proposed, notably on the choice of
proper bin size and random walk [1], obtaining effective convergence using Wang-Landau is still
challenging. Practically, differences in flatness and high dimensional effects (measure
concentration) are major hurdles for convergence.
To address these problems, we introduce two novel strategies based upon non symmetric random
walks and the exploitation by the random walk of local geometric features of the landscape.
In addition, there exists very few versatile implementations of the Wang-Landau algorithm. We
provide such an implementation, in C++, decoupling all key ingredients of the algorithm
(physical system, data structures, flat histogram rule, etc). The code, which is to be released in
the Structural Bioinformatics Library [3], is used to obtain results on peptides and a model
protein (BLN69), whose energy landscapes have been studied elsewhere [4,5]. Convergence
speedups of several orders of magnitude are obtained.
[1] Bornn, Jacob, Del Moral and Doucet
An adaptive interacting Wang--Landau algorithm for automatic density exploration
J. of Computational and Graphical Statistics, 22 (3), 2013
[2] Chevallier and Cazals
Exploiting geometric features of the Potential Energy Landscape improves the convergence
speed of Wang-Landau
Submitted, 2017.
[3] The Structural Bioinformatics Library: modeling in biomolecular science and beyond
Cazals and Dreyfus
Bioinformatics, 7 (33), 2017
[4] Hybridizing rapidly growing random trees and basin hopping yields an improved exploration
of energy landscapes
Roth, Dreyfus, Robert and Cazals
J. of Computational Chemistry, 37 (8), 2016.