Identifying repeated structural elements in folded proteins

A previously described method for partitioning amino acid dihedral angle space generates a set of cluster centers which may be used to represent residue backbones in three dimensional protein structures (R.T. Miller et al., 1993). 99 non-homologous polypeptide chains are extracted from Brookhaven Protein Data Bank files and abstracted into sequences of these cluster centers, then this data set is examined for repeated sub-sequences. Repeated sequences of amino acid conformational clusters indicate duplicated motifs or sub-structures in folded proteins, independent of amino acid sequence identity. This representation is significantly more detailed than standard secondary structure descriptors (helix, strand, turn, etc.) and facilitates the identification of common structural elements without regard to sequence length. The work presented examines the degree to which three example proteins share sub-structures with the remainder of the data set (ribonuclease A, erythrocruorin, and actinozanthin), the specification of /spl alpha/-helices and /spl beta/-strands under this system, and the sensitivity of the method to visual differences between structures.<>