Computer Science ETDs

Author

Torin Adamson

Publication Date

5-1-2016

Abstract

Molecular docking drives many important biological processes including immune system recognition and cellular signalling. Molecular docking occurs when molecules interact and form complexes. Predicting how specific molecules dock with each other using computational methods has several applications including understanding diseases and virtual drug design. The goal of molecular docking prediction is to find the lowest energy ligand states. The lower the energy state, the more probable the state is docked and biologically feasible. Existing automated computational methods can be time intensive, especially when using direct molecular dynamic simulation. One way to reduce this computational cost is to use more coarse-grained models that approximate molecular docking. Coarse-grained molecular docking prediction is generally performed first by sampling ligand states using a rigid body model or a partial flexibility model to reduce computation, then by screening the states. The ligand states are screened using a scoring function, usually a potential energy function for interactions between the atoms in each molecule. Ligand state search algorithms still have a significant computational cost if a large portion of the state space is to be explored. Instead of an automated ligand state search method, a human operator can explore the state space instead. Haptic force feedback devices providing guidance based off the energy function can aid the human operator. Haptic-guidance has been used for immersive semi-automatic and manual molecular docking on a single operator scale. A large amount of ligand state space can be explored with many human operators in a crowdsourced effort. Players in an interactive crowdsourced protein folding puzzle game have aided in finding protein folding prediction solutions, but without haptic feedback. Interactive crowdsourced methods for molecular docking prediction is not well-explored, although non-interactive crowdsourced systems such as Folding@home can be adapted for molecular docking. This thesis presents a molecular docking game that produces low potential energy ligand states and motion paths with crowdsource scale potential. In an exploratory user study, participants were assigned four different types of devices with varying levels of haptic guidance to search for a potentially docked ligand state. The results demonstrate some effect on the type of device and haptic guidance seen in the study. However, differences are minimal thus potentially enabling the use of commonly available input devices in a crowdsourced setting.

Language

English

Keywords

Probabilistic Roadmap Methods, Molecular Docking, Haptics, Crowdsourcing

Document Type

Thesis

Degree Name

Computer Science

Level of Degree

Masters

Department Name

Department of Computer Science

First Committee Member (Chair)

Tapia, Lydia

Second Committee Member

Kelley, Patrick

Third Committee Member

Jacobson, Bruna

Fourth Committee Member

Castellanos, Joel

Project Sponsors

National Institutes of Health, National Science Foundation

Share

COinS