GeNVoM: Read Mapping Near Non-Volatile Memory

Area of Science:

• Genomics
• Bioinformatics
• Computer Architecture

Background:

• Next Generation Sequencing (NGS) generates vast amounts of short DNA fragments (reads).
• Read mapping, aligning these reads to a reference genome, is crucial for genomic analysis but computationally intensive.
• Existing methods struggle with the scale and noise inherent in NGS data.

Purpose of the Study:

• To develop a scalable, energy-efficient, and high-throughput solution for DNA read mapping.
• To address the limitations of traditional computing architectures for processing large genomic datasets.
• To leverage non-volatile Ternary Content Addressable Memory (TCAM) for efficient similarity matching in read mapping.

Main Methods:

• Introduction of GeNVoM, a novel accelerator designed from the ground up for read mapping.
• Co-design of algorithms and non-volatile TCAM-based hardware.
• Evaluation against an ASIC baseline for performance and energy consumption.

Main Results:

• GeNVoM achieves up to 3.67x improvement in throughput compared to an ASIC baseline.
• GeNVoM reduces energy consumption by up to 1.36x.
• The solution demonstrates high scalability and efficiency for NGS data processing.

Conclusions:

• GeNVoM offers a significant advancement in read mapping technology.
• The proposed approach enables more efficient and faster genomic research.
• This work highlights the potential of specialized hardware accelerators for bioinformatics challenges.