MICA-aligner

MICA-aligner accelerates short-read alignment for next-generation sequencing by exploiting Intel MIC architecture and vector units to perform high-throughput alignment of 150 bp paired-end reads.

Key Features:

• Optimization for Intel MIC: Tailored to the Intel MIC platform, leveraging approximately 60-core MIC cards and 512-bit vector units for vectorized operations.
• Enhanced intra-core parallelism: Maximizes intra-core parallelism to compensate for fewer cores per MIC card and to enable efficient execution across multiple MIC cards.
• Novel seeding strategy: Employs an innovative seeding strategy to improve alignment speed and accuracy for 150 bp paired-end reads.
• Performance benchmarks: Achieves a 4.9-fold speed increase over BWA-MEM on a top-end six-core CPU with one MIC card, surpasses GPU-based SOAP3-dp, and provides a 14.1-fold acceleration over BWA-MEM when using three MIC cards.
• Scalability: Scales across multiple MIC cards and nodes, demonstrated by aligning 90 whole-genome sequencing samples (17.47 terabases) within an hour on 400 nodes of the Tianhe-2 supercomputer.

Scientific Applications:

• NGS alignment on MIC-enabled systems: High-throughput alignment of next-generation sequencing data on Intel MIC-equipped servers and supercomputers.
• Large-scale whole-genome sequencing: Processing and alignment for large-scale whole-genome sequencing and other extensive genetic analyses.

Methodology:

MICA-aligner implements 512-bit vectorized operations on Intel MIC cores, maximizes intra-core parallelism, employs a novel seeding strategy, and scales across multiple MIC cards and compute nodes for alignment of 150 bp paired-end reads.