Deep reinforcement learning-based pairwise DNA sequence alignment method compatible with embedded edge devices.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Sequence alignment is crucial for understanding biological relationships.
• Cloud computing and GPUs have accelerated bioinformatics, but edge solutions are needed for low-cost healthcare.
• Developing efficient algorithms for edge devices is essential for democratizing genomic diagnostics.

Purpose of the Study:

• To present EdgeAlign, a novel deep reinforcement learning method for pairwise DNA sequence alignment on edge devices.
• To enable sequence alignment on low-power hardware, independent of sequence length.
• To optimize the algorithm for resource-constrained environments using AutoML.

Main Methods:

• Utilized deep reinforcement learning, specifically a deep Q-network (DQN) agent, for sequence alignment.
• Implemented a sliding window approach on fixed-length sub-sequences for hardware resource optimization.
• Employed AutoML for neural network model size reduction.
• Deployed the compact model on edge devices (NVIDIA Jetson Nano and Xilinx Artix-7 FPGA).

Main Results:

• Demonstrated successful pairwise DNA sequence alignment on stand-alone edge devices.
• Achieved hardware resource consumption independent of sequence length.
• Showcased the compact and efficient nature of the EdgeAlign model.
• Validated the approach using Influenza sequences from NCBI.

Conclusions:

• EdgeAlign offers a viable solution for performing DNA sequence alignment on resource-constrained edge devices.
• The method facilitates the development of affordable, portable, and efficient genomic diagnostic tools.
• This work paves the way for advanced bioinformatics analyses in low-power, edge-computing scenarios.