Abstract: Protein secondary structures mean regular patterns in natural 3D structures such as ALPHA-helix and BETA-strand and protein secondary structure prediction is to estimate them from amino acid sequences. The secondary structure prediction not only becomes the base to infer structural properties from structurally unknown proteins, but also is useful as the constraint to predict 3D structures. The trial to predict protein secondary structures has been started from 1970’s and has gradually but steadily advanced until now. Today, average prediction accuracy rate exceeds 80% and over, so it can be said the prediction becomes a reliable and practical method. Here are summarized fundamental approaches to the secondary structure prediction, their recent development and the cautionary notes for their practical use. We had compared 72 proteins of known structure from their relationship between amino acid sequences and secondary structures. In this study, we are going to propose a server implementing a method to improve the accuracy in protein secondary structure prediction. This method is completely based on the prediction result, which is obtained by the online prediction tools to have a combined prediction of higher quality.