Date of Award

Summer 2013

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

Committee Director

Nikos Chrisochoides

Committee Member

Jing He

Committee Member

Andrey Chernikov

Abstract

In the post-genomic era, proteomics research presents a new frontier in life science. Proteins play roles in virtually every biological process, and understanding their atomic structures is the key to unraveling how they carry out their work. Compared to the over half million protein sequences in UniProt, only around 25,000 unique sequences have been atomically modeled and deposited to PDB (Protein Databank). Cryoelectron Microscopy (cryoEM) is an important biophysical technique that produces 3D subnanometer resolution images of molecules not amenable to past approaches like x-ray crystallography or nuclear magnetic resonance. De novo modeling is becoming a promising approach to derive the atomic structure of proteins from the cryoEM 3D images at "medium" resolutions{between 5 and 10 A.

Distance measurement along 1D skeletons of 3D images is an important step in de novo modeling. Despite the need of such measurement, little has been investigated about its accuracy in searching for an effective method. We propose a method to refine the skeletal length via line simplification after selecting the appropriate segmentation from the density map using Hausdorff distances. Complementarily, we developed a motion planning approach to estimate the minimum length of a loop lying completely within a contour of the density map. To test the methods, loops between 1 and 10 residues in length were extracted from atomic structures in PDB and used to generate density maps at 8 A resolution, along with experimentally derived density maps from EMDB (Electron Microscopy Databank).

Share

COinS