Parallel Pairwise Correlation Computation On Intel Xeon Phi Clusters

摘要

Co-expression network is a critical technique for the identification ofinter-gene interactions, which usually relies on all-pairs correlation (orsimilar measure) computation between gene expression profiles across multiplesamples. Pearson's correlation coefficient (PCC) is one widely used techniquefor gene co-expression network construction. However, all-pairs PCC computationis computationally demanding for large numbers of gene expression profiles,thus motivating our acceleration of its execution using high-performancecomputing. In this paper, we present LightPCC, the first parallel anddistributed all-pairs PCC computation on Intel Xeon Phi (Phi) clusters. Itachieves high speed by exploring the SIMD-instruction-level and thread-levelparallelism within Phis as well as accelerator-level parallelism among multiplePhis. To facilitate balanced workload distribution, we have proposed a generalframework for symmetric all-pairs computation by building bijective functionsbetween job identifier and coordinate space for the first time. We haveevaluated LightPCC and compared it to two CPU-based counterparts: a sequentialC++ implementation in ALGLIB and an implementation based on a parallel generalmatrix-matrix multiplication routine in Intel Math Kernel Library (MKL) (alluse double precision), using a set of gene expression datasets. Performanceevaluation revealed that with one 5110P Phi and 16 Phis, LightPCC runs up to$20.6\times$ and $218.2\times$ faster than ALGLIB, and up to $6.8\times$ and$71.4\times$ faster than single-threaded MKL, respectively. In addition,LightPCC demonstrated good parallel scalability in terms of number of Phis.Source code of LightPCC is publicly available athttp://lightpcc.sourceforge.net.