[Paper Review] FASTR: Reimagining FASTQ via Compact Image-inspired Representation

Adrian Tkachenko, Sepehr Salem+8 more|arXiv (Cornell University)|Jan 23, 2026

Genomics and Phylogenetic Studies0 citations

TL;DR

FASTR introduces a lossless, 8-bit encoding that combines nucleotide and base quality into a single byte, enabling faster I/O, smaller storage, and seamless integration with existing tools without decompression. It outperforms FASTQ and related formats across multiple platforms while preserving data integrity and enabling ML-ready representations.

ABSTRACT

Motivation: High-throughput sequencing (HTS) enables population-scale genomics but generates massive datasets, creating bottlenecks in storage, transfer, and analysis. FASTQ, the standard format for over two decades, stores one byte per base and one byte per quality score, leading to inefficient I/O, high storage costs, and redundancy. Existing compression tools can mitigate some issues, but often introduce costly decompression or complex dependency issues. Results: We introduce FASTR, a lossless, computation-native successor to FASTQ that encodes each nucleotide together with its base quality score into a single 8-bit value. FASTR reduces file size by at least 2x while remaining fully reversible and directly usable for downstream analyses. Applying general-purpose compression tools on FASTR consistently yields higher compression ratios, 2.47, 3.64, and 4.8x faster compression, and 2.34, 1.96, 1.75x faster decompression than on FASTQ across Illumina, HiFi, and ONT reads. FASTR is machine-learning-ready, allowing reads to be consumed directly as numerical vectors or image-like representations. We provide a highly parallel software ecosystem for FASTQ-FASTR conversion and show that FASTR integrates with existing tools, such as minimap2, with minimal interface changes and no performance overhead. By eliminating decompression costs and reducing data movement, FASTR lays the foundation for scalable genomics analyses and real-time sequencing workflows. Availability and Implementation: https://github.com/ALSER-Lab/FASTR

Motivation & Objective

• Motivate the need for a more storage- and I/O-efficient sequencing data format beyond FASTQ.
• Propose FASTR as a lossless, computation-native successor to FASTQ.
• Show that FASTR with its header encoding and 8-bit base-quality packing reduces size and speeds up compression/decompression.
• Demonstrate compatibility and low overhead with existing tools and workflows.
• Provide an extensible, machine-learning-ready ecosystem for FASTQ–FASTR conversion and downstream analysis support.

Proposed method

• Encode each nucleotide with its base quality into a single 8-bit value via an implicit partitioned range encoding.
• Represent read metadata with a global file header that eliminates redundancy from per-read headers.
• Provide four FASTQ-to-FASTR conversion modes with varying header and data retention trade-offs.
• Use a sentinel value (255) as a read delimiter to enable lossless, streamable decoding.
• Normalize and scale base quality scores from various Phred schemes into a common numeric range for encoding.
• Integrate FASTR with downstream tools (e.g., minimap2) with minimal interface changes and no performance overhead.

Figure 1: Overview of FASTR processing modes compared to a FASTQ record

Experimental results

Research questions

• RQ1 Can a lossless, compact 8-bit encoding of base and quality achieve substantial file-size reductions without sacrificing reversibility?
• RQ2 Does FASTR enable faster generation, compression, and decompression compared to FASTQ and other formats across multiple sequencing technologies?
• RQ3 Is FASTR compatible with existing bioinformatics tools with minimal code changes?
• RQ4 How does header optimization affect data redundancy and parsing efficiency?
• RQ5 Can FASTR support machine-learning workflows by providing numerical representations directly from reads?

Key findings

• FASTR achieves at least ~2× file-size reduction over FASTQ while remaining lossless and directly usable for downstream analyses.
• General-purpose compressors on FASTR yield higher compression ratios and faster compression/decompression than on FASTQ (e.g., 2.47×–4.8× faster compression; 2.34×–1.75× faster decompression across Illumina, HiFi, ONT).
• Compared with SAM/BAM/CRAM, FASTR is faster to generate, smaller, and reference-free, with efficient conversion back to FASTQ and minimal overhead for header optimization.
• FASTR enables drop-in compatibility with minimap2 with negligible performance overhead, while reducing I/O and output sizes.
• The implicit partitioned range encoding packs base and quality into a single 8-bit value, with a reserved 255 as a delimiter for reliable, parallelizable decoding.
• Four FASTQ-to-FASTR modes provide flexible trade-offs between header compactness and data availability for different workflows.