Abstract
Simulating molecular systems is a significant use of high-performance computing. However, molecular dynamics simulations are limited by the time scales of events that can be reliably accessed. Such events are often precisely those that are of interest, such as chemical reactions or substantial conformational changes. Accelerated simulation methods can bias a simulation towards these so-called `rare events' more quickly, but they are typically computationally expensive or require the identification of low-dimensional representations that describe the event.In this thesis, improvements to simulation methods for accessing these rare events are presented. The first is the automation and generalisation of the boxed molecular dynamics (BXD) method, making it more efficient and usable with events that require higher-dimensional representation.
Additionally, a virtual reality framework for interactive molecular dynamics (iMD-VR) is presented as a strategy for the rapid identification of pathways and collective variables that can be used with existing accelerated molecular dynamics methods.
The framework is evaluated in a user study, in which it was found that VR enables a statistically significant advantage over traditional interfaces for performing tasks in molecular systems. Furthermore, the framework is evaluated for generating initial pathways on a benchmark system, alanine dipeptide, and found to produce reasonable pathways. These pathways were then optimised for use with metadynamics, an accelerated sampling method, to produce converged free energy surfaces. The framework is further tested in larger systems including knotting pathways in the hypothetical protein MJ0366 and loop motions in the enzyme cyclophilin A. In these cases, it was possible to produce the desired pathways and initial conditions with the iMD-VR framework. The use of adaptive sampling with Markov models to perform follow-up sampling on these systems is critically evaluated.
Date of Award | 7 May 2019 |
---|---|
Original language | English |
Awarding Institution |
|
Supervisor | David Glowacki (Supervisor) & Simon N McIntosh-Smith (Supervisor) |