An Algorithm to Cluster Orthologous Proteins across Multiple Genomes

In GOLD (Genomes OnLine Database), the re It has been a successful technique, for researches on genome evolution and for functional annotation of newly sequenced genomes, to construct an OPCs(Orthologous Protein Clusters) with the best reciprocal BLAST hits from multiple complete-genomes. It, however, needs time-labor processes to make the OPCs by hand and biological analysis. In order to reduce the load, we propose an automatic parallel computing method that clusters OPs(Orthologous Proteins) from multiple complete-genomes. For systematic representation of clustering OPs, a mathematical frame of a vector is suggested. The algorithm starts, on the hypercube model, with parallelism of clustering pairwise genomes(CPG) and parallelizes all the processes of clustering multiple genomes(CMG). In CPG, all pairwise- comparisons are divided into sub-pairwise- comparisons, and the local results of clustering OPs between two genomes are integrated and broadcast to each processor. In CMG, all the clustering-processes are split into sub- clustering-processes, and the local results of OPCs among multiple genomes are broadcast to each processor and integrated in it.