A HYBRID PROGRESSIVE-ITERATIVE ALGORITHM FOR MULTIPLE STRUCTURAL ALIGNMENT
By Dmitry Lupyan
Researcher, Mount Sinai School of Medicine
A new multiple structural alignment algorithm is described. Using a number of quality criteria, it is shown that the alignments produced with the new algorithm are an improvement over previous manual or automatic alignments available in two widely used databases at the family, superfamily and fold levels. Detailed analysis of the alignments obtained with several relevant folds and superfamilies indicates that MAMMOTH-mult produces functionally and evolutionary relevant trees, as well as conservation of functional motifs in the alignments. An important advance over previous approaches is in the computational cost of the structural alignments, with typical alignments taking only a median time of 5 CPU seconds in a single R12000 processor. This makes the approach particularly useful for large-scale applications in structural genomics.