Content area
Abstract
This dissertation combines rigid body motion kinematics and statistical analysis techniques to extract information from detailed dynamic simulations and large databases of biomolecular structures. This information is then used to quantify and elucidate structural patterns that could be used to design functional nano-structures or provide new targets for ligand-based drug design. In this regard, three particular classes of problems are examined.
First, we propose new methods for estimating the stiffness of continuum filament models of helical nucleic acid structures. In this work, molecular dynamics is used to sample RNA helices consisting of several base-pairs fluctuating about an equilibrium position. At equilibrium, each base-pair has a tightly clustered probability distribution and so we can describe the rigid body motion of the helix as the convolution of highly concentrated probability densities on SE(3).
Second, the structure and dynamics of a common RNA non-helical motif is classified. We examine several RNA bulges with varying sequences and helix curvature, and establish degrees of similarity (and dissimilarity) in the bulge motif according to the nucleic acid type of the bulge and surrounding base-pairs. Both the "static" X-ray-crystal and NMR structures and the dynamics generated from molecular dynamics simulations are used to quantify the flexibility and conservative aspects of the motif. The resulting classification scheme provides bulge motifs that could be included in a toolbox of "nanostructures" where one could pick the pieces to design a structure that has the needed shape and desired behavior.
Finally, we analyze a large collection of adenosine binding sites, focusing on the functional region of the binding site. We provide a new analysis tool that finds spatial patterns in adenosine binding pockets by examining the relative pose (position and orientation) between the adenosine ligand and the amino acids at each binding site. The similarities of the numerous adenosine binding pockets are calculated according to the pose similarity and homogeny of the structures. We show that correlations between the binding pockets are multifaceted and illustrate our findings using similarity plots and multiple correlation calculations for a comprehensive analysis.
