GenASM: A High-Performance, Low-Power Approximate String Matching Acceleration Framework for Genome Sequence Analysis (2009.07692v1)

Abstract: Genome sequence analysis has enabled significant advancements in medical and scientific areas such as personalized medicine, outbreak tracing, and the understanding of evolution. Unfortunately, it is currently bottlenecked by the computational power and memory bandwidth limitations of existing systems, as many of the steps in genome sequence analysis must process a large amount of data. A major contributor to this bottleneck is approximate string matching (ASM). We propose GenASM, the first ASM acceleration framework for genome sequence analysis. We modify the underlying ASM algorithm (Bitap) to significantly increase its parallelism and reduce its memory footprint, and we design the first hardware accelerator for Bitap. Our hardware accelerator consists of specialized compute units and on-chip SRAMs that are designed to match the rate of computation with memory capacity and bandwidth. We demonstrate that GenASM is a flexible, high-performance, and low-power framework, which provides significant performance and power benefits for three different use cases in genome sequence analysis: 1) GenASM accelerates read alignment for both long reads and short reads. For long reads, GenASM outperforms state-of-the-art software and hardware accelerators by 116x and 3.9x, respectively, while consuming 37x and 2.7x less power. For short reads, GenASM outperforms state-of-the-art software and hardware accelerators by 111x and 1.9x. 2) GenASM accelerates pre-alignment filtering for short reads, with 3.7x the performance of a state-of-the-art pre-alignment filter, while consuming 1.7x less power and significantly improving the filtering accuracy. 3) GenASM accelerates edit distance calculation, with 22-12501x and 9.3-400x speedups over the state-of-the-art software library and FPGA-based accelerator, respectively, while consuming 548-582x and 67x less power.

• The paper introduces a modified Bitap algorithm that overcomes quadratic time and storage limitations for efficient genome sequence alignment.
• It leverages a dual hardware design—GenASM-DC for distance calculation and GenASM-TB for traceback—to achieve up to 116x throughput and 37x energy savings.
• The framework offers practical benefits for personalized medicine, evolutionary studies, and outbreak monitoring by enabling faster, low-power genomic analyses.

GenASM: A Framework for High-Performance, Low-Power Genome Sequence Analysis

The paper "GenASM: A High-Performance, Low-Power Approximate String Matching Acceleration Framework for Genome Sequence Analysis" presents a novel approach to accelerate genome sequence analysis by employing a modified Bitap algorithm within the GenASM framework. This work is notable for addressing the computational and power limitations inherent in current genome sequencing processes, emphasizing its applicability to personalized medicine, evolutionary studies, and outbreak containment.

Core Contributions

1. Algorithmic Innovation:

GenASM enhances the Bitap algorithm to tackle the limitations associated with traditional dynamic programming-based string matching algorithms used in genome sequencing, which typically require quadratic execution time and storage resources. By adapting Bitap for genome sequences, GenASM can efficiently process long reads, owing to modifications like multi-word bitvector handling and parallel computation capabilities.

2. Hardware Acceleration:

The framework is divided into two primary hardware components—GenASM-DC for distance calculation and GenASM-TB for traceback operations. The GenASM-DC utilizes a systolic array architecture to perform rapid bitwise operations for string matching, while GenASM-TB conducts traceback to derive optimal genome alignments. This dual-component hardware design ensures processing efficiency and power conservation.

Performance and Efficiency

Numerical Results:

GenASM demonstrates formidable gains in performance and energy efficiency across various applications of genome sequencing:

• GenASM outmatches existing state-of-the-art software and hardware systems, such as BWA-MEM, Minimap2, and DARwin, by significant margins. For long reads, GenASM achieves up to 116x greater throughput compared to 12-thread Minimap2 execution, with an accompanying reduction of power consumption by up to 37 times.
• When integrated into pipelines replacing the alignment steps of BWA-MEM and Minimap2, it results in total pipeline execution time reductions of up to 6.5x for different dataset types.
• The framework is compared with FPGA-based solutions like Shouji and exhibits 3.7x better performance while improving power efficiency in pre-alignment filtering tasks.

Implications and Future Directions

Practical Implications:

The GenASM framework exemplifies a substantial advancement toward the development of efficient bioinformatics solutions capable of keeping pace with the rapidly growing data output from modern sequencing technologies. By balancing computational speed and power consumption, GenASM aligns with the increasing demands of on-site and real-time genomic investigations, such as outbreak tracking and personalized medicine.

Theoretical Impact:

The co-design of algorithms and hardware in GenASM underscores the potential for other bioinformatics applications to benefit from targeted hardware acceleration. The approach could extend beyond genomic contexts, suggesting a trend towards more specialized computational ecosystems optimized for specific algorithmic frameworks.

Speculations on Future AI Developments:

The integration of machine learning and AI-driven insights with frameworks akin to GenASM could lead to even greater optimizations and discoveries in genomics. Accelerating pattern recognition and motif discovery in large-scale DNA datasets, tailored machine learning models built on top of such efficient processing frameworks could foreseeably revolutionize personalized genomics and predictive health analytics.

Through providing this comprehensive, versatile, and efficient framework for string matching in genomics, GenASM sets a promising trajectory for future interdisciplinary endeavors in computational biology and hardware design.