Structural comparison of related molecules is an important tool for understanding protein structure and function. Several structural comparison techniques have been reported in the literature (e.g., Taylor and Orengo, Holm and Sander, Falicov and Cohen, Godzik and Skolnick). Each of these methods defines its own metric of molecular similarity, requires some type of user-definable parameter, and generates a single alignment. While these methods have many merits, the different metrics used in the methods may be difficult to interpret quantitatively, particularly when one varies the user-definable parameter values. We are investigating a structure alignment method which uses only root- mean-square distance (RMSD) as the molecular similarity metric. Rather than generating a single answer, alignments are produced for varying number of matching residues. In the simplest analysis, users can examine the maximum number of matching residues for a maximum RMSD, or the minimum RMSD when matching a fixed number of residues. Because the results are distance measures, alignments of different molecules may be compared. An algorithm that uses our simple metric has been implemented. The program employs rigid body transformations generate candidate alignments nd dynamic programming to score each candidate. The candidate with the lowest RMSD for each number of matching residues is reported, both graphically (as a MidasPlus or Chimera graphics object) and textually (as a Multiple Sequence Format file). We are currently evaluating the performance of our algorithm against the results reported in the literature.
Showing the most recent 10 out of 508 publications
