# Coordinated Meta-Storms enables comparison of million-level microbiomes

## TL;DR

This paper introduces a new method for efficiently comparing large microbiome datasets using phylogeny-based distances.

## Contribution

The novel Coordinated Meta-Storms approach enables scalable and efficient pairwise comparison of million-level microbiomes.

## Key findings

- CMS uses a dynamic GPU architecture for fast microbiome comparison.
- Million-level microbiomes can be compared within 40 hours on a single node.
- The method is implemented in CUDA, HIP, and C++ for high performance.

## Abstract

Phylogeny-based distance serves as the foundation in the evaluation of microbiome beta diversity. However, the exponential growth of data volume has recently introduced substantial computational challenges, while current methods remain limited in their scalability and efficiency.

To solve these issues, we introduce Coordinated Meta-Storms (CMS), a novel approach for large-scale microbiome phylogeny-based distance calculation. CMS incorporates a dynamic multiple Graphics Processing Unit architecture based on a self-adaptive data decomposition strategy and optimization, which empowers pairwise comparison of million-level microbiomes within 40 h on a single computing node.

Coordinated Meta-Storms is implemented by CUDA, HIP and C++. Source code is available at GitHub (https://github.com/qdu-bioinfo/Coordinated-Meta-Storms) and has been published on Zenodo (DOI: 10.5281/zenodo.15802999).