CUDASW++: optimizing Smith-Waterman sequence database searches for CUDA-enabled graphics processing units

Abstract

BackgroundThe Smith-Waterman algorithm is one of the most widely used tools for searching biological sequence databases due to its high sensitivity. Unfortunately, the Smith-Waterman algorithm is computationally demanding, which is further compounded by the exponential growth of sequence databases. The recent emergence of many-core architectures, and their associated programming interfaces, provides an opportunity to accelerate sequence database searches using commonly available and inexpensive hardware.FindingsOur CUDASW++ implementation (benchmarked on a single-GPU NVIDIA GeForce GTX 280 graphics card and a dual-GPU GeForce GTX 295 graphics card) provides a significant performance improvement compared to other publicly available implementations, such as SWPS3, CBESW, SW-CUDA, and NCBI-BLAST. CUDASW++ supports query sequences of length up to 59K and for query sequences ranging in length from 144 to 5,478 in Swiss-Prot release 56.6, the single-GPU version achieves an average performance of 9.509 GCUPS with a lowest performance of 9.039 GCUPS and a highest performance of 9.660 GCUPS, and the dual-GPU version achieves an average performance of 14.484 GCUPS with a lowest performance of 10.660 GCUPS and a highest performance of 16.087 GCUPS.ConclusionCUDASW++ is publicly available open-source software. It provides a significant performance improvement for Smith-Waterman-based protein sequence database searches by fully exploiting the compute capability of commonly used CUDA-enabled low-cost GPUs.

Accelerating Phylogeny-Aware Short DNA Read Alignment with FPGAs

“…Several alternative implementations for accelerating the Smith-Waterman algorithm using FPGAs ( [14], [15]), vec- tor operations on x86 CPUs [16], and GPUs (e.g., using CUDA [17]) exist. However, because the PaPaRa alignment kernel differs significantly from the standard SmithWaterman implementation, we omit a more detailed review at this point.…”

An Overview of Hardware-Based Acceleration of Biological Sequence Alignment

“…Device Database Performance searched (Liu, Schmidt, Voss, Schroder & Muller-Wittig, 2006) GTX 7800 Swiss-Prot 650 MCUPS (Liu, Huang, Johnson & Vaidya, 2006) GTX 7800 983 protein 241 MCUPS sequences (Manavski & Valle, 2008) GTX 8800 Swiss-Prot 1.9 GCUPS (Liu et al, 2009) GTX 280 Swiss-Prot 9.5 GCUPS (Liu et al, 2010) GTX 280 Swiss-Prot 17 GCUPS (Kentie, 2010) GTX 275 Swiss-Prot 21.4 GCUPS …”

“…Furthermore, it is shown to scale almost linearly with the amount of GPUs used by simply splitting up the database. Various improvements have been suggested to the approach presented in (Manavski & Valle, 2008), as shown in (Akoglu & Striemer, 2009;Liu et al, 2009). In (Liu et al, 2009), for sequences of more than 3,072 amino acids an 'inter-task parallelization' method similar to the systolic array and OpenGL approaches is used as this, while slower, requires less memory.…”

“…Various improvements have been suggested to the approach presented in (Manavski & Valle, 2008), as shown in (Akoglu & Striemer, 2009;Liu et al, 2009). In (Liu et al, 2009), for sequences of more than 3,072 amino acids an 'inter-task parallelization' method similar to the systolic array and OpenGL approaches is used as this, while slower, requires less memory. The 'CUDASW++' solution presented in (Liu et al, 2009) manages a maximum speed of about 9.5 GCUPS searching Swiss-Prot on a Geforce GTX 280 graphics card.…”

“…In (Liu et al, 2009), for sequences of more than 3,072 amino acids an 'inter-task parallelization' method similar to the systolic array and OpenGL approaches is used as this, while slower, requires less memory. The 'CUDASW++' solution presented in (Liu et al, 2009) manages a maximum speed of about 9.5 GCUPS searching Swiss-Prot on a Geforce GTX 280 graphics card. An improved version, 'CUDASW++ 2.0' has been published recently (Liu et al, 2010).…”