# Accelerating minimap2 for whole-genome alignment

## TL;DR

This paper introduces mm2-plus, an optimized version of the minimap2 aligner that significantly speeds up whole-genome alignment without losing accuracy.

## Contribution

The novel contribution is the fine-grained parallel chaining algorithm and improved chain differentiation in mm2-plus.

## Key findings

- mm2-plus accelerates genome alignment by 1.6× to 7.2× for human, plant, and primate genomes.
- The method maintains alignment accuracy while improving CPU utilization and reducing idle threads.
- Optimizations are effective for megabase-scale sequences typical in modern genome assemblies.

## Abstract

Recent advances in long-read sequencing and genome assembly techniques have enabled the generation of high-quality assemblies, often comprising megabase-scale sequences that span entire chromosomes. This results in longer but fewer sequences per genome, which affects the parallelization efficiency of whole-genome alignment tools. Current methods that assign one thread per query sequence now face suboptimal CPU use and longer runtimes because the processing of fewer sequences leaves many threads idle. We present mm2-plus, a fast and efficient method for whole-genome alignment, built upon the commonly used minimap2 aligner. Our improvements include a fine-grained parallel chaining algorithm and a fast method for differentiating primary and secondary chains. These optimizations accelerate the alignment of human, plant, and primate genomes by 1.6× to 7.2× without compromising accuracy.

Source code is available at https://github.com/at-cg/mm2-plus and https://doi.org/10.5281/zenodo.18220923.