Qi Yutao, Lin Feng
BioInformatics Research Center, School of Computer Engineering Nanyang Technological University, Nanyang Avenue, Singapore 639798.
Cell Mol Biol Lett. 2005;10(2):281-5.
Retrieving homologous DNA and protein sequences from existing databases is a fundamental routine in bioinformatics research. Programs of the NCBI BLAST family are widely used for this purpose. We evaluated paraBLAST, a parallelised version of the NCBI BLAST algorithm, using a Message Passing Interface (MPI) on a multi-node compute cluster. Here, we propose static and dynamic database-partitioning schemes based on the availability of the cluster. We evaluated the application of the algorithm in querying nucleotide sequences against a large-scale sequence database with different numbers of database partitions, and hence, different numbers of CPUs. Since the program's tasks are performed independently of each other, each available CPU can run its own copy of BLAST queries, resulting in reduced interference between processes and leading to a highly scalable solution.
从现有数据库中检索同源DNA和蛋白质序列是生物信息学研究的一项基本常规操作。NCBI BLAST家族的程序广泛用于此目的。我们在多节点计算集群上使用消息传递接口（MPI）评估了paraBLAST，它是NCBI BLAST算法的并行版本。在此，我们根据集群的可用性提出了静态和动态数据库分区方案。我们评估了该算法在针对具有不同数量数据库分区（即不同数量CPU）的大规模序列数据库查询核苷酸序列中的应用。由于程序的任务彼此独立执行，每个可用CPU都可以运行自己的BLAST查询副本，从而减少进程之间的干扰并产生高度可扩展的解决方案。
