XSW 2.0 - A fast Smith-Waterman Algorithm Implementation on Intel Xeon Phi Coprocessors.

Introduction

XSW is a new parallel Smith-Waterman algorithm for searching large-scale biological sequence databases on the Xeon Phi-based neo-heterogenous architecture.

Evaluations on real protein sequence databases(Non-Redundant protein + UniProtKB/TrEMBL protein, around 36GB) shows that XSW 2.0 performance of 100 GCUPS on a hybrid CPU-Xeon Phi platform(1x Xeon Phi 7110p + 1x E5-2620).

Downloads

The latest binary file:
zip file.
tar.gz file.
We use Intel compiler 2013 SP1 to compiler the source code.Please check your intel library version.

Installation and Usage

Three files must be provided to run XSW: the scoring matrix, query sequence and the database. The query sequence and database should be in FASTA format.
For example, steps to run the query testRefSeq0511.txt to the database db.env_nr with the default gap penalties and scoring matrix are as following:

For native version:

Step 1: Preproccess the database:
./makemicdb.out db.env_nr 244
[NOTE:] if your Xeon Phi has 61 cores, it is 244, if 60 cores it is 240.
Step 2: Upload the database and program to xeon phi:
scp db.env_nr*.* blosum62.mat testRef* XSW mic0:
Step 3: Login xeon phi card:
ssh mic0
Step 4: Run XSW on xeon phi:
./XSW blosum62.mat testRefSeq0511.txt db.env_nr

For offload version:

Step 1: Preproccess the database:
./makemicdb4.1.out db.env_nr
Step 2: Run XSW:
./XSW blosum62.mat testRefSeq0511.txt db.env_nr

Papers

XSW: Accelerating Biological Database Search on Xeon Phi
Lipeng Wang, Yuandong Chan, Xiaohui Duan, Haidong Lan, Xiangxu Meng, and Weiguo Liu
The 28th IEEE International Parallel & Distributed Processing Symposium (IPDPS 2014), Phoenix,USA, May, 2014.

Support or Contact

If you have any questions, please contact:Weiguo,Liu ( weiguo.liu@sdu.edu.cn).