The Potential of Intel Xeon Phi for DNA Sequence Analysis

Genetic information is increasing exponentially, doubling every 18 months. Analyzing this information within a reasonable amount of time requires parallel computing resources. While considerable research has addressed DNA analysis using GPUs, so far not much attention has been paid to the Intel Xeon Phi coprocessor. In this paper we present an algorithm for large-scale DNA analysis that exploits the thread-level and the SIMD parallelism of the Intel Xeon Phi coprocessor. We evaluate our approach for various numbers of cores and thread allocation affinities in the context of real-world DNA sequences of mouse, cat, dog, chicken, human and turkey. The experimental results on Intel Xeon Phi show speed-ups of up to 10× compared to a sequential implementation running on an Intel Xeon processor E5.