Lossy Compression of Quality Values in Sequencing Data.

Area of Science:

• Genomics
• Bioinformatics
• Data Compression

Background:

• The rapid expansion of human DNA sequencing generates vast amounts of data, creating significant storage challenges.
• Compression techniques are crucial for managing large-scale genomic datasets, particularly SAM files used in DNA alignment.

Purpose of the Study:

• To investigate the effectiveness of lossy compression techniques for quality values in SAM files.
• To analyze the impact of these lossy compression methods on the CRAM file format and single nucleotide polymorphism (SNP) calling.

Main Methods:

• Experiments were conducted using the NA12878 dataset with varying fold coverages.
• A novel lossy compression model, dynamic binning, was introduced and compared against existing methods (Illumina binning, LEON, QVZ).
• Compression ratios for CRAM files and SNP calling performance were evaluated.

Main Results:

• Lossy compression techniques significantly improve the compression ratio of CRAM files.
• The evaluated lossy techniques did not negatively impact SNP calling accuracy.
• In some instances, lossy compression even demonstrated a potential to boost SNP calling performance.

Conclusions:

• Lossy compression is a viable strategy for reducing the storage footprint of genomic data in CRAM format.
• The application of lossy compression for quality values in SAM files offers benefits without compromising essential downstream analyses like SNP calling.