H-BLAST

H-BLAST accelerates protein sequence alignment by parallelizing BLASTP and BLASTX computations across heterogeneous CPU and GPU systems to enable faster searches of large biological sequence databases.

Key Features:

• Parallel Processing: Executes alignments in parallel on systems with multiple CPUs and GPUs to increase throughput.
• Locally Decoupled Seed-Extension Algorithm: Implements a locally decoupled seed-extension algorithm optimized for GPU architectures.
• Performance Tuning Mechanism: Includes a performance-tuning mechanism to adjust operations across different CPU/GPU combinations.
• Compatibility with NCBI-BLAST: Supports BLASTP and BLASTX core functionalities and produces alignment results identical to NCBI-BLAST.
• Significant Speed Improvements: Reports 4–10× speedups for NCBI-BLASTP and 5–7.2× for NCBI-BLASTX versus sequential NCBI-BLAST, and surpasses a 16-threaded NCBI-BLASTX when run with two CPU threads and two GPUs.
• Superiority Over GPU-BLAST: Demonstrates 1.5–4× faster performance than GPU-BLAST in tested configurations.

Scientific Applications:

• Comparative Genomics: Enables rapid large-scale sequence comparisons required for comparative genomics analyses.
• Functional Annotation of Genes: Accelerates BLASTP/BLASTX searches used in protein function annotation workflows.
• Evolutionary Biology: Facilitates evolutionary analyses that rely on extensive protein sequence alignments.
• Large-Scale Sequence Analysis: Supports high-throughput processing of large genomic and proteomic sequence datasets.

Methodology:

Optimizes the BLAST algorithm for heterogeneous computing by parallel execution on CPUs and GPUs, employing a locally decoupled seed-extension algorithm tailored to GPU architectures and dynamically adjusting operations based on hardware configuration.