Combat Search rapidly for highly similar proteincoding sequences using bipartite graph mat.pdfVIP

COMBAT: Search Rapidly For Highly Similar Protein-Coding Sequences Using Bipartite Graph Matching Bing Sun, Jacob T. Schwartz, Ofer H. Gill, and Bud Mishra Courant Institute of Mathematical Sciences, New York University 251 Mercer Street, New York, NY 10012, USA. bingsun@, jack@, gill@, mishra@ Abstract. Comparing vertebrate genomes requires efficient cross-species sequence alignment programs. We describe COMBAT, a new mer-based method which can search rapidly for highly similar translated genomic se- quences, with the stable-marriage algorithm with incomplete lists (SMI) as a filter scheme. We apply the COMBAT program to the comparative analysis of the human with the most recent bovine genome assemblies, and 84%∼95% of the homologous blocks identified by this program are confirmed by BLASTZ. 1 Introduction In the past decade many genome projects have produced complete genomes for increasingly many organisms. Since 1999 many tools have proven effective in aligning large genomic sequences of two closely related organisms. These include MUMmer [4], GLASS [1], AVID [2], DIALIGN [8], LAGAN [3], BLASTZ [9], BLAT [7], and etc. Common characteristics in many of these programs are: i) they assume the conserved regions of the sequences being aligned appear in the same order and orientation, as is particularly likely for closely related organisms; ii) they build tables of scores for matches and mismatches between amino acids or nucleotides incorporating penalties for insertions or deletions, and from these constructs obtain mathematically ‘optimal’ alignments; iii) many local alignment programs search for exact or spaced exact matches, and then extend the local similarities in both directions in passes directed by specified scor