Efficient Construction of Multiple Geometrical Alignments for the Comparison of Protein Binding Sites

We proceed from a method for protein structure comparison in which information about the geometry and physico-chemical properties of such structures are represented in the form of labeled point clouds, that is, a set of labeled points in three-dimensional Euclidean space. Two point clouds are then compared by computing an optimal spatial superposition. This approach has recently been introduced in the literature and was shown to produce very good similarity scores. It does not, however, establish an alignment in the sense of a one-to-one correspondence between the basic units of two or more protein structures. From a biological point of view, alignments of this kind are of great interest, as they offer important information about evolution, heredity, and the mutual correspondence between molecular constituents. In this paper, we therefore developed a method for computing pairwise or multiple alignments of protein structures on the basis of labeled point cloud superpositions.