A VECTOR PROJECTION METHOD FOR PREDICTING SUPERSECONDARY MOTIFS

For the 11 types of most frequently occurring supersecondary motifs, w e used a new method-the vector projection method-to predict a protein' s supersecondary structure. In a training set of peptides and a test s et of peptides we obtained a satisfactory result, with a prediction ac curacy of about 90%. The high prediction accuracy indicates that this method is reasonable for predicting the folding motifs of proteins. Th is work provides insight into the problem of predicting a protein's lo cal structure accurately, and is of particular value in protein modeli ng, prediction, and molecule design.