SWIFOLD: Smith-Waterman implementation on FPGA with OpenCL for long DNA sequences

Area of Science:

• Bioinformatics
• Computational Biology
• Computer Science

Background:

• The Smith-Waterman (SW) algorithm is crucial for sequence similarity searches but computationally intensive.
• High computational demands limit SW algorithm applicability in certain scenarios.
• Parallel architectures like GPUs and FPGAs are explored to accelerate large-scale sequence alignment workloads.

Purpose of the Study:

• To present and evaluate SWIFOLD, a novel parallel Smith-Waterman implementation on FPGAs using OpenCL.
• To assess SWIFOLD's performance and resource utilization across various kernel configurations.
• To compare SWIFOLD against state-of-the-art methods for DNA sequence alignment.

Main Methods:

• Development of SWIFOLD, a parallel Smith-Waterman algorithm optimized for FPGAs.
• Utilized OpenCL for programming the FPGA architecture.
• Evaluated performance using diverse datasets and compared with existing implementations, including GPU-based solutions.

Main Results:

• SWIFOLD demonstrates superior average performance for small and medium DNA sequence datasets.
• Performance achieved by SWIFOLD is independent of input size and sequence similarity.
• SWIFOLD exhibits competitive performance against the latest GPU generations for large datasets.

Conclusions:

• SWIFOLD emerges as a viable and affordable solution for accelerating Smith-Waterman alignment of DNA sequences of any size.
• The FPGA-based implementation achieves significant computational speeds, reaching up to 270 GCUPS.
• SWIFOLD offers a promising alternative for efficient large-scale biological sequence analysis.