Classification of amino acids based on comparative analysis of contacts in DNA-protein complexes and specific DNA-protein interactions

We propose a classification of amino acid residues based on the events of contact formation between particular residues and DNA nucleotides, i.e., using the most integral properties that characterize interactions organizing DNA-protein complexes. We apply the Voronoi-Delaunay tessellation to draw statistics of contacts and of contact areas for a set of 1937 DNA-protein complexes. Similarity of amino acid residues is defined upon comparison of corresponding rows and matrices of contacts and areas of contacts. Nine measures of distance have been used to estimate the closeness of rows. Residues have been grouped by three hierarchical and two nonhierarchical clustering methods. In a total tree built using nine metrics with three hierarchical methods, we show that clustering centers (pairs of amino acids) in the main groups are always constant while other relationships between objects vary. Major classes of up to six amino acids correspond to certain local structures of the polypeptide chain. These data can be taken into account when designing DNA-protein ligands.