Based on the concept that the structural class of a protein is mainly determined by its secondary structure sequence, a new algorithm for prediction of the structural class of a protein is proposed. By use of the number ofα -helices, β -strands, and βαβ fragments, the structural class of a protein can be predicted by an algorithm based on the increment of diversity (ID), in which the sole prediction parameter—the increment of diversity is used as the index of prediction of structural class of a protein. The results indicate that the high rates of correct prediction are obtained for complete set (standard set) from Brookhaven Protein Data Bank-CD ROM (PDB) published in October 1995 and the test set newly released from Brookhaven Protein Data Bank-CD ROM (PDB) before July 1998, respectively.