Computational Challenges of Next Generation Sequencing Pipelines Using Heterogeneous Systems

Abstract

We are rapidly entering the era of genomics. The dramatic cost reduction of DNA sequencing due to the introduction of Next Generation Sequencing (NGS) techniques has resulted in an exponential growth of genetics data. The amount of data generated, and its associated processing into useful information, poses serious computational challenges. Here, we give a brief introduction of NGS, show a typical NGS processing pipeline, and show the associated challenges from a computational perspective. A case study is presented where one component of the NGS processing pipeline is accelerated: BWA-MEM, the de-facto industry-standard for the mapping stage. This is a first step in achieving a fully heterogeneously accelerated NGS pipeline.