[Paper Review] BioSEAL: In-Memory Biological Sequence Alignment Accelerator for Large-Scale Genomic Data

Roman Kaplan, Leonid Yavits+1 more|arXiv (Cornell University)|Jan 17, 2019

Genomics and Phylogenetic Studies62 references4 citations

TL;DR

BioSEAL proposes a non-von Neumann, in-memory computing accelerator based on resistive content addressable memory (RCAM) to enable massively parallel biological sequence alignment for large-scale genomics. It achieves up to 57x speedup and 156x better energy efficiency than state-of-the-art solutions by performing associative processing directly on genomic databases, significantly outperforming traditional dynamic programming and heuristic methods like BLAST.

ABSTRACT

Genome sequences contain hundreds of millions of DNA base pairs. Finding the degree of similarity between two genomes requires executing a compute-intensive dynamic programming algorithm, such as Smith-Waterman. Traditional von Neumann architectures have limited parallelism and cannot provide an efficient solution for large-scale genomic data. Approximate heuristic methods (e.g. BLAST) are commonly used. However, they are suboptimal and still compute-intensive. In this work, we present BioSEAL, a Biological SEquence ALignment accelerator. BioSEAL is a massively parallel non-von Neumann processing-in-memory architecture for large-scale DNA and protein sequence alignment. BioSEAL is based on resistive content addressable memory, capable of energy-efficient and high-performance associative processing. We present an associative processing algorithm for entire database sequence alignment on BioSEAL and compare its performance and power consumption with state-of-art solutions. We show that BioSEAL can achieve up to 57x speedup and 156x better energy efficiency, compared with existing solutions for genome sequence alignment and protein sequence database search.

Motivation & Objective

• To address the computational bottleneck of large-scale biological sequence alignment in genomics.
• To overcome the limitations of von Neumann architectures in handling the massive parallelism required for dynamic programming algorithms like Smith-Waterman.
• To design a hardware accelerator that enables energy-efficient, high-throughput sequence alignment using in-memory associative processing.
• To outperform existing heuristic (e.g., BLAST) and exact alignment methods in both performance and energy efficiency.
• To demonstrate the feasibility and advantages of resistive content addressable memory (RCAM) for biological sequence database search.

Proposed method

• The architecture employs resistive content addressable memory (RCAM) to perform associative processing, enabling simultaneous comparison of query sequences against entire genomic databases.
• A novel associative processing algorithm is designed to map the Smith-Waterman dynamic programming computation onto the RCAM-based in-memory architecture.
• The system performs content-based parallel matching, where each RCAM cell stores a genomic sequence segment and evaluates similarity on-the-fly.
• The design leverages the inherent parallelism of RCAM to execute multiple alignment computations in a single cycle, avoiding data movement bottlenecks.
• The architecture is optimized for energy efficiency by minimizing data movement and utilizing in-memory computation.
• The method supports both DNA and protein sequence alignment by adapting the similarity scoring and matching logic to biological sequence types.

Experimental results

Research questions

• RQ1 Can an in-memory, non-von Neumann architecture based on RCAM achieve significant performance and energy efficiency gains for biological sequence alignment?
• RQ2 How does the proposed associative processing algorithm map the dynamic programming computation of Smith-Waterman onto RCAM for full database alignment?
• RQ3 What is the achievable speedup and energy efficiency improvement of BioSEAL compared to traditional von Neumann systems and heuristic tools like BLAST?
• RQ4 To what extent can RCAM-based processing eliminate data movement overhead in large-scale sequence alignment?
• RQ5 Can the proposed accelerator handle both DNA and protein sequence databases with high throughput and low latency?

Key findings

• BioSEAL achieves up to 57x speedup over state-of-the-art solutions for genome sequence alignment.
• The accelerator demonstrates 156x better energy efficiency compared to existing approaches for protein sequence database search.
• The RCAM-based architecture enables massively parallel associative processing, drastically reducing the time required for full database alignment.
• The proposed associative algorithm effectively maps the computationally intensive Smith-Waterman algorithm onto the in-memory hardware, minimizing latency.
• BioSEAL outperforms heuristic methods like BLAST in both accuracy and performance, while maintaining superior energy efficiency.
• The system is scalable and applicable to both DNA and protein sequence alignment, showing promise for large-scale genomics applications.