If letters representing the amino acid sequences of related proteins are lined up you can see patterns where a similar amino acid appears in most or, even better, in all of the sequences at some positions. These positions are called "conserved." At other positions in the sequences different amino acids occur across the alignment. These are called "unconserved" positions The particular pattern of conserved and unconserved positions can give an additional clue to the structure of the protein at a position.
We have had some success using an alignment based prediction method. Our prediction of the structure of Protein-Tyrosine Phosphatase, for example, had more than 75% of the amino acids predicted correctly. Automated methods which take a similar approach have now been developed by Burkhard Rost and Chris Sander in Germany. This level of accuracy may be good enough to build a complete three dimensional model of a protein.
