Par-eXpress: A tool for analysis of sequencing experiments with ambiguous assignment of fragments in parallel

Abstract

With new high-throughput and low-cost sequencing technologies, an increasing amount of genetic data is becoming available to researchers. While the analysis of this vast amount of data has great potential for future scientific advances, it becomes imperative to exploit parallelism in order to process this data efficiently. In this paper, we address probabilistic assignment of ambiguously mapped fragments. This is a very significant, but time consuming, process for downstream analysis of genomic data. We develop a distributed-memory parallel version of a popular probabilistic fragment assignment tool, eXpress. In our experiments, we show that our approach achieves significant speedups over eXpress without decreasing its accuracy. The speedup we achieve increases as the number of iterations and/or data size increases.