A sphere-based descriptor for matching protein structures

This paper describes the use of a descriptor based on the number of alpha-carbon atoms within a sphere centered on the alpha-carbon in each amino acid residue in a protein. The descriptor can be used instead of the residue types in a dynamic programming algorithm, thus providing an efficient way of aligning protein structures. The method is applied to the alignment of protein families and to database searching. The results indicate that the method can quickly align protein sequences considering the 3D structure and can find proteins that are 3D structurally similar and dissimilar to the target protein.