grailbio

grailbio implements AF4, an alignment-free fusion detector that identifies fusion fragments in cell-free nucleic acid (cfNA) sequencing data to enable sensitive detection of fusion events despite cfNA-specific challenges.

Key Features:

• Alignment-Free kmer-Based Analysis: AF4 uses an alignment-free, kmer-based approach to identify candidate fusion fragments.
• Two-Stage Processing — Stage One: The tool decomposes each fragment into kmers and maps them to corresponding genes to identify potential fusion pairs.
• Two-Stage Processing — Stage Two: A max-cover criterion is applied to filter candidate fragments, reducing false positives while preserving true fusion events.
• Detection Modes and Validation: AF4 supports both targeted and de novo fusion detection modes and has been validated on benchmark simulated and real RNA-seq data as well as clinical and cell-line cfNA datasets.
• cfNA-Specific Handling: The method addresses high sequencing depth, low allele fractions, short fragment lengths, and the use of unique molecular identifiers in cfNA sequencing.

Scientific Applications:

• Cancer Research and Diagnostics: Detection of fusion genes from cfNA sequencing to inform tumor biology and enable non-invasive cancer detection.
• Targeted and Exploratory Fusion Analysis: Both targeted assays and de novo discovery of fusion events in cfNA and RNA-seq datasets.

Methodology:

Alignment-free kmer-based analysis; decomposition of fragments into kmers and mapping to genes to identify candidate fusion pairs; application of a max-cover criterion to filter candidate fragments; support for targeted and de novo detection modes.