[Prediction of protein conformation using a doublet code method]

It is suggested that regions of irregular structure, beta-structure, and alpha-helix are composed of 2, 3, and 5 amino acid residue long elements (structurons), respectively, and that the structurons are encoded solely by residue pairs (doublet codons) (i, i + 1), (i, i + 2), (i, i + 4), respectively. Tables of codons are obtained by statistical analysis of the data on the distribution of these pairs in available secondary structures of 62 proteins. These tables are used to obtain distributions of t-, beta- and alpha-codons for an amino acid sequence of protein. When codons of different structures superpose, that is, include the same sequence regions, selection is performed, the selection being performed so to obtain as much as possible number of the non-superposed codons of different structures. The distributions of structurons obtained after this selection are used for localization of structurons in the sequence and prediction of secondary structure on the basis of this localization. The prediction method is illustrated. An accuracy of the method has been tested on the basis an casual selection of fifteen proteins and found equal 64% for secondary structure on the whole and 79%, 53%, 61% for alpha-helix, beta-structure and coil respectively. This result is similar or better than that communicated for contemporary methods.