GPU accelerated adaptive banded event alignment for rapid comparative nanopore signal analysis

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Nanopore sequencing offers portable, real-time genetic analysis for diagnostics and field applications.
• Efficient bioinformatic algorithms are essential for processing raw nanopore signal data.
• Adaptive Banded Event Alignment (ABEA) is critical for sequence polishing and identifying modified nucleotides like DNA methylation.

Purpose of the Study:

• To parallelize and optimize the ABEA algorithm for efficient execution on heterogeneous CPU-GPU architectures.
• To accelerate the analysis of nanopore sequencing data for improved real-time applications.

Main Methods:

• Optimized memory usage, computations, and load balancing between CPU and GPU.
• Implemented a parallelized version of the ABEA algorithm, named f5c.
• Tested performance against existing CPU-only ABEA implementations.

Main Results:

• The f5c implementation achieved a ~3-5x speedup compared to the optimized CPU-only ABEA in Nanopolish.
• Demonstrated on-the-fly DNA methylation detection using f5c on a GPU-equipped System on Chip (SoC).

Conclusions:

• Complex genomic analyses are feasible on lightweight computing systems.
• The optimized ABEA algorithm (f5c) benefits both portable sequencing applications and High-Performance Computing (HPC).
• Source code for f5c and GPU-optimized ABEA is publicly available.