f5c

f5c performs GPU-accelerated adaptive banded event alignment of raw Oxford Nanopore signal data as a reimplementation of Nanopolish's call-methylation module to enable DNA methylation detection.

Key Features:

• Adaptive Banded Event Alignment (ABEA): f5c implements the ABEA dynamic programming algorithm to align raw nanopore ionic current events to a reference sequence and detect modified bases such as methylated cytosines.
• GPU Acceleration: f5c parallelizes and optimizes ABEA for heterogeneous CPU–GPU architectures, yielding approximately 3–5× faster performance compared to the CPU-only Nanopolish implementation.
• Memory and Load Balancing: The implementation applies optimized memory usage and computational load distribution between CPUs and GPUs to maximize throughput.
• Real-Time Analysis: Computational efficiency supports on-the-fly DNA methylation detection from nanopore signal streams.
• Embedded and SoC Compatibility: f5c is implemented to run efficiently on GPU-equipped embedded systems and System on Chip (SoC) architectures.
• CUDA Support: f5c supports CUDA acceleration to exploit NVIDIA GPUs.

Scientific Applications:

• DNA Methylation Detection: Identification of methylated cytosine bases in basecalled Oxford Nanopore reads for epigenetic studies and biomarker discovery.
• Portable Sequencing and Field Applications: Rapid comparative nanopore signal analysis for on-site environmental monitoring and clinical diagnostic scenarios requiring immediate processing.
• High-Performance Computing: GPU optimization facilitates more efficient use of HPC resources for large-scale genomic analyses.

Methodology:

f5c employs a parallelized implementation of the Adaptive Banded Event Alignment (ABEA) algorithm optimized for CPU–GPU execution, with strategic memory management and computational load distribution between CPUs and GPUs and support for CUDA acceleration on NVIDIA GPUs.