My undergraduate thesis is on computational approaches to protein structure problems. In particular, I analyzed the geometry of proteins using the Voronoi tessellation and tried to make sense of that data via unsupervised machine learning (clustering) and data visualization techniques. My adviser was Darrell Schroeter in the Reed College physics department.
I demonstrate an elastic lists inspired way of exploring the statistics of protein structure. Several interactive histograms display the multidimensional dataset and also allow users to iteratively refine queries in such a way that they will always find matches.
A protein’s distance matrix is the n by n symmetric matrix that gives the pairwise distances between a protein’s amino acid residues. Often one chooses some cutoff distance to form a binary contact map, which simply tells whether residues i and j are “in contact”. There are several ways to define the distance between residues (considering the alpha carbon, sidechain centroid, or shared Voronoi edge), and no general agreement on contact map cutoff distances. This application allows one to compare contact maps for different distance measures and cutoff parameters.
2010 May 12: Draft submitted to library (6 MB).
2010 April 30: Draft submitted for defense (6 MB).
2010 April 25: Now with 324% more transitions! (3 MB).
2010 April 11: First draft release (2 MB).
2010 March 31: Slides from thesis talk (3 MB).