Hyper-BLAST : a parallelized BLAST on cluster systems

Abstract

Basic Local Alignment Search Tool (BLAST) is an important similarity search tool in bioinformatics. It is used to find biologically similar sequences to the given query sequence from the database of the annotated sequences. For high throughput processing of huge number of query sequences, which are generated by automatic DNA sequencing machines, there have been many studies on parallel batch processing of sequence similarity search using BLAST. As the number of sequences in the database increases at an exponential rate, the search speed of BLAST becomes the most critical performance parameter. Although a parallel version of BLAST, which exploits intra-search level parallelism, has been developed by National Center for Biotechnology Information (NCBI) at National Institute of Health (NIH) and Washington University, it is targeted only on symmetric multiprocessors (SMP) machines. Since SMP machine has a limitation on the number of processors due to its architecture, the speedup of BLAST that can be achieved solely on one SMP machine also faces inevitable performance limitation. Hence, the speedup improvement of BLAST on the SMP machine is not sufficient to cope with processing a large amount of sequences that are newly added into the database at an exponential rate. In order to use more processors for more speedup of BLAST, we consider PC cluster system as an alternative to SMP machine. This dissertation presents $\emph{Hyper-BLAST}$, a parallelized BLAST on cluster system, which can provides scalable speedup in terms of response time. $\emph{Hyper-BLAST}$ adds inter-node parallel execution techniques to the intra-search level parallelism that is used by NCBI BLAST. Logical partitioning of database is used to prepare and enable the inter-node parallel execution of intra-search level parallelism. In our parallelized BLAST on cluster system, the master node drives remote nodes to search similar sequence from logically partitioned database and collects minimal b...