Implementation of Short Read Alignment Algorithm in OpenCL on Xeon Phi Coprocessor

Aligning sequencing reads to a reference genome is often essential in many comparative genomics pipelines. With the maturation of next-generation DNA sequencing (NGS) technologies, an enormous amount of sequence data has been generated, this calls for the development of faster read alignment programs. In this paper we present an OpenCL implementation of the short read aligner BarraCUDA [1], which is based on BWA software package [2] and utilises NVIDIA´s GPGPU implementation, namely Compute Unified Device Architecture (CUDA) to parallelise the alignment of sequence reads. We re-implemented BarraCUDA aligner in OpenCL and test it on Intel´s MIC architecture device - Xeon Phi 5110P coprocessor. Experiment results show that our OpenCL aligner performs better than its CPU-based software counterpart, with a 4 times performance boost compared to 8-thread BWA.