A Two-Level Scheme for Quality Score Compression.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomic Data Analysis

Background:

• Quality score compression is crucial for managing large genomic datasets.
• Existing methods are either lossy, saving resources but losing original data, or lossless, requiring significant storage.
• Bioinformaticians need efficient methods that support both preliminary analysis and in-depth data investigation.

Purpose of the Study:

• To introduce QScomp, a novel, space-efficient hierarchical representation for quality scores.
• To enable efficient lossy compression for routine analyses while preserving the ability to recover original quality scores.
• To evaluate the performance and space usage of QScomp compared to existing compression schemes.

Main Methods:

• Developed a novel decomposition of quality scores into tuples.
• Implemented a hierarchical compression strategy with separate lossy and lossless components.
• Compressed the first dimension using a lossy scheme and the second dimension for score recovery.
• Evaluated downstream analysis performance and total space usage on real genomic data.

Main Results:

• QScomp achieves competitive downstream analysis performance using only 0.49 bits per quality score on average for the lossy component.
• The total space usage, including the compressed second dimension for score recovery, is comparable to existing lossless schemes.
• The method successfully allows users to work with compressed quality scores while retaining access to original data.

Conclusions:

• QScomp provides an effective solution for balancing computational resource management and data fidelity in genomic analyses.
• This hierarchical compression approach offers a practical advantage for bioinformaticians handling large-scale sequencing data.
• QScomp demonstrates that efficient lossy compression can coexist with the capability for lossless data recovery.