Shape Representation and Invariant Description of Protein Tertiary Structure in Applications to Shape Retrieval and Classification

Each protein tertiary structure is represented by three sorts of geometric models, which are polyline curves, triangulated surfaces and volumetric solids. Moment invariants are employed to describe the shapes of the three kinds of protein models and form a multidimensional feature vector for each model. We further analyze the influence of the three representations in applications to protein shape retrieval and classification.