Structural pattern matching for functional annotation of proteins
Chen, Brian Yuan
Kavraki, Lydia E.
Master of Science
Characterizing the function of every protein is one of the most important goals of genomics. However, the expense of experimentation makes the functional "annotation" of every protein simply infeasible. Thus, techniques in comparative analysis have popularized by helping to reduce overall experimental needs through classification. Here we present two pattern matching techniques to classify proteins if they contain the same functional structures. We use the Evolutionary Trace (ET) to identify residues of a functional site for a "motif" of labeled points embedded in three dimensions (3D). Then, given any protein structure, we determine if the motif exists within it, using an adaptation of Geometric Hashing (GH). Because GH is incompatible with some aspects of ET, we then present Match Augmentation: a novel improvement on GH which improves sensitivity and performance. Our success on both randomized and biological examples demonstrates that our tools can efficiently and robustly empower biologists studying functional annotation.
Molecular biology; Computer science; Biology