Partitioned multi-MUM finding for scalable pangenomics with MumemtoM

Area of Science:

• Computational Biology
• Bioinformatics

Background:

• Pangenome datasets are rapidly expanding, requiring efficient alignment methods.
• Existing methods struggle to scale with hundreds of high-quality genomes.

Purpose of the Study:

• To develop scalable and efficient methods for constructing pangenome alignments.
• To enhance the Mumemto tool for incorporating new genome assemblies.

Main Methods:

• Introduced MumemtoM (Mumemto Merge) with two novel partitioning and merging strategies.
• Implemented parallel, memory-efficient, and updateable computation of maximal unique matches (multi-MUMs).
• Developed a string-based merging strategy for phylogenetic-tree-aware alignment.

Main Results:

• MumemtoM successfully scales to 474 human haplotypes, a feat unmatched by other multi-MUM methods.
• The new strategies enable highly parallel and memory-efficient multi-MUM computation.
• Introduced a time-memory tradeoff for adaptability to diverse computational environments.

Conclusions:

• MumemtoM significantly advances scalable pangenome alignment construction.
• The tool is efficient, adaptable, and capable of handling large-scale genomic data.
• Phylogenetic-tree-aware merging offers novel insights for evolutionary genomics.