nBLAST-JC: An accelerated deployment of nucleotide BLAST on high-performance cluster
Sita Rani, Aman Kataria, O. P. Gupta, Pankaj Bhambri, Hong Min
Abstract
BLAST is the most practiced algorithm to perform the task of sequence alignment. To cope with the voluminous growth in molecular data and reduce processing time, several accelerated versions of BLAST have been proposed by researchers. In this paper, an expeditious design of nucleotide BLAST named nBLAST-JC for a high-performance cluster (HPC) is proposed. The proposed algorithm is implemented on the Hadoop framework using JCuda. It can run multiple query sequences in parallel. The proposed nBLAST-JC gave a speed-up ranging from 7.1 × to 9 × in comparison to HS-BLASN and 1.8 × to 2.3 × in comparison to G-BLASTN (‘blastn’ mode). This algorithm is highly scalable for huge-sized databases due to the fundamental characteristics of the Hadoop framework. The performance of nBLAST-JC can be further enhanced using available CPU cores along with the GPU of each client node. The reliability of the proposed model can further be improved using a two-server approach.