Mathematical methods for protein structural motif recognition

Abstract

In this thesis, two algorithms for protein structural motif recognition are presented. A program is described which successfully recognizes the occurrence of the right-handed parallel 8-helix fold from protein sequence data alone. When run on sequences of unknown structure the program identifies the fold in a significant number of microbial outer membrane and cell-surface proteins implicated in infectious disease. A search algorithm is introduced for unsupervised discovery of protein sequence patterns that are significantly correlated with structural motifs. The algorithm identifies known, biologically relevant patterns as well as several potentially novel motifs. It is hoped that both algorithms may contribute to a better understanding of the connection between protein sequence and structure.