Toward an Embedded SoC for Mobile DNA Sequencing Applications.

Area of Science:

• Genomics and Bioinformatics
• Computer Engineering
• Hardware Acceleration

Background:

• Portable DNA sequencing technologies are advancing rapidly.
• Current portable sequencers require bulky external systems for bioinformatics analysis.
• There is a need for integrated, efficient computing solutions for mobile genomics.

Purpose of the Study:

• To propose and evaluate an embedded System-on-Chip (SoC) architecture for mobile DNA sequencing.
• To develop specialized hardware accelerators for key bioinformatics tasks.
• To demonstrate the feasibility of on-device bioinformatics processing.

Main Methods:

• Designed a heterogeneous SoC leveraging the RISC-V architecture.
• Incorporated two dedicated accelerators: one for deep learning-based basecalling and another for sequence alignment.
• Implemented the basecalling accelerator on an FPGA and evaluated its performance and energy efficiency.

Main Results:

• The basecalling accelerator achieved a ~2000x speedup and superior energy efficiency compared to CPUs and GPUs.
• The sequence alignment accelerator demonstrated significant performance-per-joule advantages over existing platforms.
• A preliminary ASIC prototype operated at 250 MHz with low power consumption (34 mW).

Conclusions:

• The proposed SoC architecture offers a viable solution for integrated bioinformatics in mobile DNA sequencing.
• Hardware acceleration is crucial for achieving high performance and energy efficiency in portable genomic applications.
• This work represents a significant step towards fully embedded bioinformatics hardware for ubiquitous genomic analysis.