Leon

Leon compresses next-generation sequencing (NGS) data using a reference-free probabilistic de Bruijn Graph approach to reduce storage requirements for downstream bioinformatics analyses.

Key Features:

• Reference-free compression: Constructs a probabilistic de Bruijn Graph from sequencing reads without relying on external reference genomes.
• Probabilistic de Bruijn Graph storage: Stores the de Bruijn Graph efficiently using a Bloom filter for compact representation.
• Read encoding as graph paths: Encodes each read as a path through the graph by recording an anchoring k-mer and the sequence of bifurcations encountered.
• Lossy quality-score transformation: Performs a lossy transformation on sequencing quality scores to increase compression while preserving information needed for downstream analyses.
• Compression performance: Outperforms traditional compressors such as gzip on high-throughput sequencing datasets, with reported reductions exceeding 20× on a Caenorhabditis elegans whole-genome dataset.
• Targeted data types: Tailored for high-throughput NGS outputs including whole genome, exome, RNA-seq, and metagenomics datasets.

Scientific Applications:

• Whole-genome sequencing compression: Applied to WGS datasets to substantially reduce storage footprint (e.g., >20× on a Caenorhabditis elegans dataset).
• Exome sequencing compression: Used to compress targeted exome sequencing reads while retaining analyzable sequence information.
• RNA-seq data compression: Applied to transcriptomic sequencing datasets to decrease storage and transmission costs.
• Metagenomics data compression: Employed on complex metagenomic sequencing datasets where reference-free approaches are advantageous.

Methodology:

Constructs a probabilistic de Bruijn Graph from sequencing reads without external references and stores the graph using a Bloom filter; encodes reads as paths by recording an anchoring k-mer and the sequence of bifurcations, and applies a lossy transformation to quality scores to improve compression.