Nonlinear Correlations of Protein Sequences and Symmetries of Their Structures

We investigate the nonlinear correlations of protein sequences by using the nonlinear prediction method developed in nonlinear dynamical theory. It is found that a lot of protein sequences show strong nonlinear correlations and have deterministic structures. Further investigations show that the strong nonlinear correlations of these protein sequences are due to the symmetries of their tertiary structures. Furthermore, the correlation lengths of the sequences are related to the degrees of the symmetries. These results support the duplication mechanism of protein evolution and also reveal one aspect how amino acid sequences encode their spatial structures.