SWCuda

SWCuda implements GPU-accelerated Smith-Waterman local alignment for protein sequences using the NVIDIA CUDA platform to enable high-sensitivity database searches.

Key Features:

• CUDA Implementation: Uses NVIDIA CUDA to execute the Smith-Waterman algorithm on GPUs, targeting the G80 architecture and providing direct access to GPU hardware primitives.
• Performance Optimization: Achieves speeds exceeding 3.5 GCUPS on a workstation with two GeForce 8800 GTX graphics cards, substantially outperforming traditional CPU-based Smith-Waterman implementations.
• Scalability and Efficiency: Leverages GPU acceleration to perform large-scale sequence alignments efficiently and reports speedups of approximately 2–30× over existing solutions on commodity hardware.
• Exact Smith-Waterman Implementation: Implements the exact dynamic-programming Smith-Waterman algorithm rather than heuristic approaches, preserving alignment sensitivity.

Scientific Applications:

• Protein and DNA Database Searches: Facilitates similarity searches within large protein and DNA databases where exhaustive, sensitive alignments are required.
• Exact Alignment Solutions: Provides exact local alignments as an alternative to heuristic methods such as FASTA and BLAST, maintaining higher sensitivity.

Methodology:

Implements the Smith-Waterman dynamic programming algorithm on GPUs via NVIDIA CUDA, distributing cell updates across multiple GPU cores to accelerate computation while preserving exact local-alignment scoring.