Alignment algorithm for homology modeling and threading

A DNA/protein sequence comparison is a popular computational tool for molecular biologists. Finding a good alignment implies an evolutionary and/or functional relationship between proteins or genomic loci. Sequential similarity between two proteins indicates their structural resemblance, providing a practical approach for structural modeling, when structure of one of these proteins is known. The first step in the homology modeling is a construction of an accurate sequence alignment. The commonly used alignment algorithms do not provide an adequate treatment of the structurally mismatched residues in locally dissimilar regions. We propose a simple modification of the existing alignment algorithm which treats these regions properly and demonstrate how this modification improves sequence alignments in real proteins.