A High-Efficiency SoC for Next-Generation Mobile DNA Sequencing

TL;DR

The paper addresses the bottleneck of limited onboard processing in handheld DNA sequencers by designing a 22-nm SoC that integrates a RISC-V core with dedicated accelerators for Viterbi-based sequence detection. It introduces two accelerators, AccelA for trellis construction and AccelB for both trellis construction and traceback, and demonstrates substantial performance and energy efficiency gains (up to $13\times$ faster and ~$3000\times$ more energy-efficient) over conventional platforms. Using predictive nanopore $3$-MER models and chunked event processing, the system achieves up to $2.6$ Mevents/s at $200$ MHz and supports edge processing with low overhead energy, enabling real-time basecalling within the device. The results highlight significant implications for IoT and edge-enabled genomic sensing, as embedded processing reduces data bandwidth to remote processors and enables downstream mapping/variant analysis in edge-cloud frameworks.

Abstract

Hand-sized Deoxyribonucleic acid (DNA) sequencing machines are of growing importance in several life sciences fields as their small footprints enable a broader range of use cases than their larger, stationary counterparts. However, as currently designed, they lack sufficient embedded computing to process the large volume of measurements generated by their internal sensory system. As a consequence, they rely on external devices for additional processing capability. This dependence on external processing places a significant communication burden on the sequencer's embedded electronics. Moreover, it also prevents a truly mobile solution for sequencing in real-time. Anticipating next-generation machines that include suitably advanced processing, we present a System-on-Chip (SoC) fabricated in 22-nm complementary metal-oxide semiconductor (CMOS). Our design, based on a general-purpose reduced instruction set computing (RISC-V) core, also includes accelerators for DNA detection that allow our system to demonstrate a 13X performance improvement over commercial embedded multicore processors combined with a near 3000X boost in energy efficiency.

Figures (15)

• Figure 1: An example of miniature DNA sequencer within an IoT genomic sensing system.
• Figure 2: Component overview of a possible future miniature DNA sequencing system within which more substantial SoC computing is embedded.
• Figure 3: Viterbi trellis construction algorithm.
• Figure 4: A trellis representation and its construction via state progression.
• Figure 5: The traceback algorithm.