Parallelizing bioinformatics applications with MapReduce.PDF

1 Parallelizing bioinformatics applications with MapReduce Massimo Gaggero, Simone Leo, Simone Manca, Federico Santoni, Omar Schiaratura, Gianluigi Zanetti CRS4, Edificio 1, Sardegna Ricerche, Pula, Italy Abstract—Current bioinformatics applications require both structural genomics and genome-wide association studies like management of huge amounts of data and heavy computation: EIGENSTRAT[5], PLINK[6], fastPHASE[7] and MACH[8] – fulfilling these requirements calls for simple ways to implement since they typically contain a mixture of these two computa- parallel computing. MapReduce is a general-purpose paralleliza- tional flavors. tion technology that appears to be particularly well adapted to this task. Here we report on its application, using its open source implementation Hadoop, to two relevant algorithms: II. IMPLEMENTATION BLAST and GSEA. The first is characterized by streaming computation on large data sets, while the second requires a A. Hadoop multi-pass computational strategy on relatively small data sets. The analysis of these algorithms is relevant to a wide class of Hadoop[2] offers an open source implementation of the complex applications, e.g., structural genomics and genome-wide MapReduce framework and a distributed file system (HDFS, association studies, since they typically contain a mixture of these two computational flavors. Our results are very promising and Hadoop Distributed File System) on which the framework can indicate that the frame