Alignment-free filtering for cfNA fusion fragments

Area of Science:

• Bioinformatics
• Genomics
• Molecular Biology

Background:

• Cell-free nucleic acid (cfNA) sequencing presents unique challenges for fusion detection.
• Existing methods struggle with high sequencing depth, low allele fractions, short fragments, and specialized barcodes like unique molecular identifiers (UMIs).

Purpose of the Study:

• To develop an improved method for detecting gene fusions in cfNA sequencing data.
• To address the limitations of current fusion detection tools in handling complex cfNA characteristics.

Main Methods:

• Introduced AF4, an alignment-free, k-mer based method for sensitive and rapid candidate fusion fragment detection.
• Implemented a max-cover criterion for filtering spurious matches and retaining authentic fusion fragments.
• AF4 supports both targeted and de novo fusion detection modes.

Main Results:

• AF4 achieves high sensitivity and is significantly faster than existing tools.
• The max-cover criterion effectively reduces false positives while preserving true fusion signals.
• Demonstrated AF4's performance on simulated, RNA-seq, and clinical cfNA data.

Conclusions:

• AF4 offers a robust and efficient solution for gene fusion detection in cfNA sequencing.
• The tool's speed and accuracy make it suitable for analyzing challenging cfNA datasets.
• AF4 is open-sourced and available for broader research applications.