Document Type

Technical Report

Department

Computer Science and Engineering

Publication Date

2006-01-01

Filename

wucse-2006-48.pdf

Technical Report Number

WUCSE-2006-48

Abstract

NCBI BLASTP is a popular sequence analysis tool used to study the evolutionary relationship between two protein sequences. Protein databases continue to grow exponentially as entire genomes of organisms are sequenced, making sequence analysis a computationally demanding task. For example, a search of the E. coli. k12 proteome against the GenBank Non-Redundant database takes 36 hours on a standard workstation. In this thesis, we look to address the problem by accelerating protein searching using Field Programmable Gate Arrays. We focus our attention on the BLASTP heuristic, building on work done earlier to accelerate DNA searching on the Mercury platform. We analyze the performance characteristics of the BLASTP algorithm and explore the design space of the seed generation stage in detail. We propose a hardware/software architecture and evaluate the performance of the individual stage, and its effect on the overall BLASTP pipeline running on the Mercury system. The seed generation stage is 13x faster than the software equivalent, and the integrated BLASTP pipeline is predicted to yield a speedup of 50x over NCBI BLASTP. Mercury BLASTP also shows a 2.5x speed improvement over the only other BLASTP-like accelerator for FPGAs while consuming far fewer logic resources.

Comments

Permanent URL: http://dx.doi.org/10.7936/K74F1NZR

Download

Share

COinS