Prediction of surface and interior regions in proteins--Part I: Linear tripeptide sequences identify structural boundaries in proteins.

Previously, using an IBM or compatible program called SurfacePlot, it was shown that amino acid parameters of hydrophilicity, accessibility and flexibility could successfully predict linear accessible or surface regions observed in the x-ray structure of proteins. In addition, these surface regions defined the boundaries of secondary structure regions in proteins. In this paper we show that tripeptides containing the polar residues G K S D E N P T A R and Q also predict surface regions of proteins. Although tripeptides containing these residues predicted 91% of the observed accessible regions, the combination of polar tripeptide predictions with SurfacePlot predictions significantly reduced overpredictions (36% to 17%) and improved accuracy from 71% to 83% while maintaining an 86% prediction rate of observed x-ray-defined linear accessible regions. This new program also predicted 86% of beta-turn regions. The data indicated that not all beta-turns are highly exposed to solvent and that beta-turns should not be used as the sole criteria for defining surface regions.