Prediction of surface and interior regions in proteins--Part II: Predicting secondary structure in regions bound by surface exposed regions.

A new secondary structure prediction algorithm (SecondaryPlot) is described and has been developed for the IBM or compatible computer. This algorithm successfully predicts the boundaries of secondary structural regions by combining hydrophilicity, accessibility and flexibility parameters. Within the limits of these predicted structural regions, five algorithms (Chou and Fasman; Garnier, Osguthorpe and Robson; Delange and Roux; Barkovsky and Bandarin; and Lim) were used to predict helix and sheet regions. All of the algorithms correctly predicted 50-60% of the x-ray defined regions. Predictions were improved by combining the two most successful algorithms of Lim and Chou and Fasman. By combining the boundary predictions (hydrophilicity, accessibility, flexibility and linear polar tripeptide sequences) with secondary structure predictions (Lim, Chou and Fasman), approximately 80% of the x-ray defined structural regions were predicted correctly for 30 proteins (86% for the alpha-class, 79% for the beta-class and 82% for the alpha/beta class of proteins). In our hands, the best of the five algorithms studied to predict the state (helix or sheet) of these structural regions predicted 62% of helical regions and 64% of extended regions in a test set of fifteen proteins. However, SecondaryPlot predicted 78% of helical regions and 71% of extended regions for the same test. SecondaryPlot provides a significant improvement to structure prediction and is a necessary first step in the tertiary structure prediction of proteins.