4.3 Prediction of Secondary Structure
A model of the unknown protein can be constructed using the amino acid sequence alignment and the backbone structure of the known structure. Insertions or deletions very often occur between regular secondary structural regions (e.g., loops and otherwise unordered structure). Structure prediction (modelling) based on sequence homology should become increasingly more powerful as more protein structures are determined. Evidence to support this outlook comes from the fact that there appears to be a limited number of tertiary folds (estimates range from <1000 to <8000) as judged from the 80 different folds already represented in the protein structure database (Orengo et al., 1994). They estimate that we currently know the structure of about 1% of all naturally occuring superfamilies.
Recently, Rost & Sander (1993) have reported an algorithm which uses evolutionary information contained in multiple sequence alignments as input to neural networks. The authors claim > 70% accuracy in the prediction of three classes of secondary structure (helix, sheet, and other) if the protein contains at least one known homologoue. These authors have also implemented an automated mail server for protein secondary structure prediction using this method PredictProtein server.
Generated with CERN WebMaker