WFA-GPU: gap-affine pairwise read-alignment using GPUs.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Genomic sequencing technologies produce long, noisy reads requiring efficient alignment.
• Classical dynamic programming alignment methods are computationally intensive for long sequences.
• The Wavefront Alignment (WFA) algorithm offers improved efficiency but requires parallelization for modern hardware.

Purpose of the Study:

• To develop a GPU-accelerated tool (WFA-GPU) for fast and accurate sequence alignment.
• To leverage the parallel processing power of GPUs for WFA algorithm implementation.
• To enable efficient alignment of long and noisy sequencing reads.

Main Methods:

• Implementation of the WFA algorithm on graphics processing units (GPUs).
• Development of a CPU-GPU co-design for parallel sequence alignment.
• Algorithmic adaptations and performance optimizations for GPU architecture.
• Utilizing a succinct WFA-data representation for efficient GPU computation.

Main Results:

• WFA-GPU achieves significant speedups, outperforming multi-threaded WFA by up to 4.3x (exact) and 18.2x (heuristic).
• WFA-GPU is up to 29x faster than other GPU aligners and up to four orders of magnitude faster than CPU aligners.
• WFA-GPU is the first GPU solution capable of accurately aligning long reads on commodity hardware.

Conclusions:

• WFA-GPU provides a highly efficient solution for aligning long and noisy sequencing reads.
• The GPU-accelerated approach significantly enhances the scalability and speed of genomic data analysis.
• WFA-GPU democratizes high-performance sequence alignment for researchers with commodity GPUs.