Date of Award

May 2020

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Chemistry

Committee Member

Steven J Stuart

Committee Member

Leah B Casabianca

Committee Member

Brian Dominy

Committee Member

Andrew Robb

Abstract

Molecular dynamics (MD) is a widely used tool to study molecular systems on atomic level. However, the timescale of a traditional MD simulation is typically limited to nanoseconds. Thus many interesting processes that occur on microseconds or larger timescale can't be studied. Hyperdynamics provides a way to extend the timescale of MD simulation. In hyperdynamics, MD is performed on a biased potential then corrected to get true dynamics provided certain conditions are met. Here, we tried to study potassium channel conductance using the hyperdynamics method with a bias potential constructed based on the potential of mean force of ion translocation through the selective filter of a potassium ion channel. However, when MD was performed on this biased potential, no ion translocation events were observed. Although some new insights were gained into the rate-limiting steps for ion mobility in this system from these negative results, no further studies are planned with this project.

The second project is based on the assumption that hybrid human{computational algorithm is more efficient than purely computational algorithm itself. Such ideas have already been studied by many \crowd-sourcing" games, such as Foldit [1] for the protein structure prediction problem, and QuantumMoves [2] for quantum physics. Here, the same idea is applied to cluster structure optimization. A virtual reality android cellphone app was developed to study global optimization of Lennard-Jones clusters with both computational algorithm and hybrid human{computational algorithm. Using linear mixed model analysis, we found statistically significant differences between the expected runtime of both methods, at least for cluster of certain sizes. Further analysis of the data showing human intelligence weakened the strong dependence of the efficiency of the computational method on cluster sizes. We hypothesis that this is due to that humans are able to make large moves that allows the algorithm to cover a large region in the potential energy surface faster. Further studies with more cluster sizes are needed to draw a more complete conclusion. Human intelligence can potentially be integrated into more advanced optimization technique and applied to more complicated optimization problems in the future. Patterns analysis of human behaviors during the optimization process can be conducted to gain insights of mechanisms and strategies of optimization process.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.