|
Structure alignment methods are useful for organizing and classifying known protein structures. Most of the work on comparing proteins has focused however on protein sequences. Given a separable scoring function, the optimal alignment between two protein sequences can always be found by dynamic programming. Finding the optimal structure alignment between two proteins is harder because the rotation and translation of one of the two structures with respect to the other must be calculated in addition to the alignment itself. Although an (approximate) optimal solution can be found (see the work of Kolodny and Linial, also part of this grant), it is an expensive computation that can only be applied to a pair of proteins, and is unrealistic for whole database comparisons. To circumvent this problem, we are developing a new approach to protein structure alignment that first translates the protein structures into linear sequences, referred to as the "structural sequences" of the protei.ns, and then uses the optimal tools for sequence alignment to obtain the optimal alignment.