CURC: a CUDA-based reference-free read compressor.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• High-throughput sequencing (HTS) generates massive datasets, posing storage and transfer challenges.
• Existing data compression tools primarily rely on Central Processing Units (CPUs), limiting efficiency for large-scale HTS data.
• The increasing availability of Graphics Processing Units (GPUs) presents an opportunity for accelerated data processing.

Purpose of the Study:

• To develop a Graphics Processing Unit (GPU)-accelerated compressor for high-throughput sequencing (HTS) data.
• To address the limitations of CPU-based compression methods for large HTS datasets.
• To leverage GPU computing power for efficient and fast sequencing data compression.

Main Methods:

• Introduced CURC, a GPU-accelerated, reference-free read compressor for FASTQ files.
• Implemented a GPU-CPU heterogeneous parallel scheme utilizing the CUDA library.
• Employed a pseudogenome approach for efficient lossless compression of DNA sequences.

Main Results:

• CURC achieves a 2-6 fold speedup in compression compared to state-of-the-art reference-free compressors.
• The compressor demonstrates competitive compression rates.
• The GPU-accelerated approach significantly enhances the efficiency of handling large HTS data.

Conclusions:

• CURC provides a powerful and efficient solution for compressing high-throughput sequencing data.
• The GPU-accelerated design makes CURC suitable for managing the data deluge in genomics.
• CURC offers a viable alternative to CPU-bound compression tools, improving processing times.