Sequence-to-graph alignment based copy number calling using a network flow formulation

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Copy number (CN) variation influences phenotypic differences and is crucial for disease association and genome assembly.
• Traditional CN calling relies on linear reference genomes, struggling with complex genomic structures and rearrangements.
• Existing graph-based methods for CN prediction often overlook graph topology, leading to inconsistencies.

Purpose of the Study:

• To introduce Floco, a novel method for copy number calling on genome graphs.
• To leverage network flow and integer linear programming for accurate CN estimation within graph structures.
• To improve upon existing read depth-based CN calling methods, especially for complex genomes.

Main Methods:

• Floco utilizes a network flow formulation applied to genome graphs.
• It calculates raw CN probabilities per graph node using the Negative Binomial distribution and base pair coverage.
• Integer linear programming is employed to compute CN flow across the entire graph.

Main Results:

• Floco demonstrated up to a 43% increase in CN prediction accuracy compared to read depth estimation alone.
• The method was tested on diverse datasets, including HiFi and ONT reads across three different graphs.
• High concordance (up to 93.2%) was achieved between predictions from multiple sequence sources.

Conclusions:

• Floco offers a significant advancement in copy number calling for genome graphs.
• The network flow approach addresses limitations of traditional methods and enhances prediction accuracy.
• Floco provides a robust tool for genomic analyses involving complex structural variations and graph-based representations.