SWIPE

SWIPE performs rapid Smith-Waterman local sequence alignments on amino acid and nucleotide databases using SIMD parallelization to increase sensitivity over heuristic methods.

Key Features:

• Algorithm: Implements the Smith-Waterman local alignment algorithm with SIMD parallelization for optimal local alignment scoring.
• Parallelization Strategy: Compares residues from sixteen different database sequences simultaneously to a single query residue, differing from Farrar's 2007 'striped' SIMD approach.
• Performance: Reaches 106 billion cell updates per second (GCUPS) on a dual Intel Xeon X5650 six-core system using a 375-residue query, reporting over sixfold speedup versus the striped approach and approximately 2.5-fold speedup with a single thread.
• Comparative Scoring Matrix Performance: For short queries with the BLOSUM50 matrix SWIPE is about twice as fast as BLAST, while with BLOSUM62 BLAST is approximately twice as fast as SWIPE.
• Hardware Instruction Set: Employs SIMD optimizations using SSSE3-capable processors.

Scientific Applications:

• Homology detection: Identifying homologous amino acid or nucleotide sequences using high-sensitivity Smith-Waterman alignments.
• Sequence variation analysis: Precisely characterizing local sequence variations and alignments.
• Large-scale database searches: Enabling more extensive database scans and exploration of larger datasets through increased alignment throughput.
• Genomics, proteomics and evolutionary studies: Supporting accurate local sequence comparisons required in genomics, proteomics, and evolutionary biology.

Methodology:

Implements the Smith-Waterman algorithm with SIMD-based parallelization that compares sixteen database sequences' residues against a query residue using SSSE3 SIMD instructions, contrasted with Farrar's 2007 'striped' approach.