Parallelization of MIRA Whole Genome and EST Sequence Assembler

The genome assembly problem is to generate the original DNA sequence of the organism from a large set of short overlapping fragments. MIRA is an open source assembler based on the Overlap Layout Consensus (OLC) graph model which addresses the assembly problem and is widely used by biologists [1,2]. Like other assemblers MIRA takes a long time to compute the assembly for large number of sequences. For example it takes around 24 hours to assemble a dataset with 1.4 million DNA sequence fragments and takes even longer for EST assemblies [3]. In this paper, we report our efforts in parallelizing MIRA assembler. The task of parallelizing MIRA assembler is challenging as it has critical code segments which are inherently sequential but are crucial for a good quality assembly. Initially, we focus on the two time consuming kernels of the MIRA assembler: graph construction, and edge weight calculation algorithm. Our implementation of the parallel code on a machine with 2 Intel Xenon X5550 Nehalem-EP quad core processors achieves linear speedup for these kernels.