ANALYSIS OF 3-DIMENSIONAL PROTEIN IMAGES

A fundamental goal of research in molecular biology is to understand p rotein structure. Protein crystallography is currently the most succes sful method for determining the three-dimensional (3D) conformation of a protein, yet it remains labor intensive and relies on an expert's a bility to derive and evaluate a protein scene model. In this paper, th e problem of protein structure determination is formulated as an exerc ise in scene analysis. A computational methodology is presented in whi ch a 3D image of a protein is segmented into a graph of critical point s. Bayesian and certainty factor approaches are described and used to analyze critical point graphs and identify meaningful substructures, s uch as a-helices and beta-sheets. Results of applying the methodologie s to protein images at low and medium resolution are reported. The res earch is related to approaches to representation, segmentation and cla ssification in vision, as well as to top-down approaches to protein st ructure prediction.