Autotuned parallel I/O for highly scalable biosequence analysis

Abstract

In recent years, the rate of genomics sequence generation increased dramatically due to significant advances in the sequencing technology. The genomics data is accumulating at an exponential rate in various databases all around the world and rapid analysis techniques will enhance the knowledge discovery in the fields of medicine and biotechnology. Analysis of such growing sequence databases demands tremendous computational power that can only be provided by massively parallel computers. Improving the performance and scalability of bioinformatics tools thus becomes a critical step in the quest to transform ever-growing raw genomics data into biological knowledge. In this paper we describe an efficient parallel implementation of a profile hidden Markov models (profile HMMs) code used for protein domain identification, along with auto-tuned parallel I/O optimization. Experimental results show linear speedup with increasing numbers of computing cores on a supercomputer, allowing the domain identification of millions of proteins in few minutes using hundreds of thousands computing cores.