MUMmerGPU

MUMmerGPU performs ultra-fast exact sequence alignments by leveraging NVIDIA's Compute Unified Device Architecture (CUDA) on commodity GPUs to execute multiple query sequences in parallel against a reference stored as a suffix tree, accelerating analysis of reads from sequencing technologies such as Solexa/Illumina, 454, and Sanger.

Key Features:

• GPU acceleration (CUDA): Executes alignment computations on commodity NVIDIA GPUs using NVIDIA's Compute Unified Device Architecture (CUDA).
• Parallel query execution: Processes multiple query sequences concurrently against a single reference to increase throughput.
• Suffix tree reference representation: Stores the reference sequence as a suffix tree for exact alignment searches.
• Exact alignment component: Implements exact sequence alignment functionality comparable to the exact alignment component of MUMmer.
• Performance improvements: Achieves more than a 10-fold speedup versus serial CPU implementations and reduces total application time by 3.5× compared to the exact alignment component of MUMmer for contemporary sequencing reads.
• Memory-efficient on GPUs: Handles memory-intensive alignment workloads on relatively low-cost GPUs.
• Support for sequencing read types: Demonstrated on reads from sequencing technologies including Solexa/Illumina, 454, and Sanger.

Scientific Applications:

• Genotyping: Enables high-throughput exact alignments used for genotype detection from sequencing reads.
• Genome resequencing: Facilitates rapid alignment of resequencing reads to a reference genome for variant detection.
• Metagenomics: Provides scalable alignment capacity for analyzing large metagenomic sequencing datasets.
• De novo genome assembly projects: Supplies exact alignment output useful for validating and scaffolding de novo assemblies.

Methodology:

Executes multiple query sequences in parallel on commodity NVIDIA GPUs via NVIDIA's Compute Unified Device Architecture (CUDA) against a single reference sequence stored as a suffix tree.