Waterfall: An Online Sequential Monte Carlo Strategy for Conformational Sampling
Date
2020-05-08
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Conformational sampling of physical systems, such as biomolecules, remains challenging since their energy landscapes are rugged with energetic and entropic barriers. Metastable states on a landscape are often separated by substantial energetic barriers and numerous local minima where a system can easily get trapped. Entropic barriers, arising from many possible conformations of a system, make finding the important states a needle-in-a-haystack problem. Enhanced sampling algorithms can potentially overcome both barriers, making it possible to sample the important states efficiently. However, existing algorithms come with certain limitations. We developed ``Waterfall Sampling” to overcome some of the limitations of existing algorithms. The algorithm employs a Sequential Monte Carlo strategy where it promotes exploration of the high-probability regions on the conformational landscape. It is an online method, that is, we can keep adding states indefinitely to improve statistics. We successfully tested Waterfall on a data-guided integrative structural biology problem and compared its efficiency with the existing algorithms.
Description
Keywords
Sampling algorithm, Integrative Structural Biology
Citation
Muniyat, M. I. (2020). Waterfall: An Online Sequential Monte Carlo Strategy for Conformational Sampling (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.